Do decision-making structure and sequence exist in

Decision Support Systems 74 (2015) 102–120

Contents lists available at ScienceDirect

Decision Support Systems journal homepage: www.elsevier.com/locate/dss

Do decision-making structure and sequence exist in health online social networks? Valeria Sadovykh a, David Sundaram a, Selwyn Piramuthu b,⁎ a b

Department of Information Systems and Operations Management, University of Auckland, Owen G. Glenn Building, 12 Grafton Road, Auckland, New Zealand ISOM Department, University of Florida, 351 Stuzin Hall, Gainesville, FL 32611-7169, USA

a r t i c l e

i n f o

Article history: Received 1 September 2014 Received in revised form 24 March 2015 Accepted 25 March 2015 Available online 3 April 2015 Keywords: Health online social networks Decision-making process Netnography Rational and anarchic decision making

a b s t r a c t Decision-making process has been explained through several models over the years. Among these, the rational and anarchical models have emerged as important representations of decision-making dynamics. The rational model and its variants of decision making emphasize recognized phases and sequence among them, while anarchical models focus on the lack of structure and sequence in many real-world decision-making contexts. In order to observe the existence of these phases and their sequence, it is critical to choose non-trivial situations in which the underlying dynamics of decision-making process are readily visible. To this end, we consider decisionmaking (DM) in Health Online Social Networks (HOSN) and verify the existence of recognized phases and the sequence in which these phases are reached. We use netnography to explore the potential of HOSN as a support tool for decision-making process. Our results confirm, extend, as well as challenge existing knowledge. Results confirm that HOSN support and empower users during their decision-making process in three specific key phases that include Intelligence, Design and Choice. We extend existing knowledge by suggesting two new phases in the decision making process that is integral to HOSN conversations, namely emotional support and sharing experiences. Our results challenge purely rational and anarchical models by recognizing the interweaving of anarchical decision sequences within the structure of rational decision making phases. These results have significant practical implications for the design of HOSN that support blended decision making processes by leveraging the wisdom of crowds. © 2015 Elsevier B.V. All rights reserved.

1. Introduction The making of decisions is ubiquitous across time and space and pervades every facet of our lives. While some decisions are made autonomously without much thought, the significance of the consequence (e.g., health-related) and existing constraints (e.g., time) generally dictate the extent of decision maker involvement and resources that are allocated to any given decision-making situation. The extent of involvement in the decision-making process clearly also depends on the relative importance of such decisions — for example, the flavour choice for the next chewing gum pack to be consumed most likely receives a trivial amount of decision-making resource when compared against that for a serious health-care related decision. The latter is more interesting from a decision-making perspective since the decision-maker is forced to be involved in the process, and the finer details of the process

⁎ Corresponding author. Tel.: +1(352) 392 8882. E-mail addresses: [email protected] (V. Sadovykh), [email protected] (D. Sundaram), selwyn@ufl.edu (S. Piramuthu).

http://dx.doi.org/10.1016/j.dss.2015.03.007 0167-9236/© 2015 Elsevier B.V. All rights reserved.

are more pronounced. Since health is of paramount importance, related decisions are oftentimes made with input from as many credible sources as is possible with the goal of minimizing risk while simultaneously improving the odds of better outcome. Moreover, health care decisions are complex by nature that necessitates a concomitant increase in information needs [6]. There is clearly a surge in interest among researchers and practitioners on issues associated with healthcare and recent explosion of social media-related activities as is evident from recent publications. As noted by Ellingsen and Monteiro [17], an integrated healthcare information system that seamlessly incorporates and delivers relevant information is necessary for improving healthcare delivery performance. Regardless of advances in related technology, when faced with a decision-making situation, it is not unreasonable to assume that the core processes that the stakeholders use to make decisions may not necessarily be disparate across domains. The involvement intensity is bound to vary depending on its perceived significance to the decision maker. However, from a research perspective, it is necessary to find evidence in reality to confirm the existence of a common core, albeit their social/behavioural nature. Health-related issues bring forth the nuances of the decision-making process that can be readily observed.

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

With the explosion of online social networks (OSN) and the potential wealth of information contained therein, we consider OSN as a support tool for decision-making (e.g., [16,69]). We study health-related decision-making situations that involve interactions with online social networks that are especially relevant for digital natives (e.g., [70]). Our choice of the health domain in OSN is primarily driven by its significance to decision makers, and therefore the higher revelation probability of finer decision-making details. The classical literature on decision-making models alludes to the existence of rational and sequential as well as anarchical processes in traditional settings. We suspect that the dynamics may not necessarily be the same in health online social network (HOSN) environments that incorporate the participation of online ‘advisors,’ since this moderating effect could potentially disrupt decision-making structure and sequence. We therefore study the influence of online ‘advisors’ on decision-making in HOSN environments. Specifically, we consider decision-making dynamics in health online social networks (HOSN) to (a) determine what decision-making stages are supported by HOSN, (b) identify any construct(s) that may not have been identified before in this context, and (c) verify if either rational or anarchical decisionmaking model is followed, with the exclusion of the other. On a related note, we are also interested in identifying the biases and strengths [59] of the human psyche that could be attenuated or enhanced through appropriate design of HOSN [34]. To operationalize the study, we use Netnography [37] to observe, elicit, and understand the problems and requirements of HOSN support for decision-making. Specifically, we evaluate the existence of the five decision-making phases (discussed in Section 2) and the sequence in which these phases are reached, if any. The rest of the paper is organized as follows: We discuss necessary background and related literature in Section 2. We discuss the research problems, requirements and associated questions in Section 3. We then narrow our focus to HOSN and discuss our results and findings from our netnographic analysis in Section 4. We conclude the paper with a brief discussion on the contributions of this study in Section 5.

2. Background and related literature We provide necessary background in this section to understand and appreciate the need for this study. To this end, we begin with a brief discussion on the essence of decision-making processes. We follow this with discussion on the different phases that are deemed to exist in decision-making processes. We then introduce online social networks, with specific emphasis on health online social networks and how it relates to stakeholder decision support.

2.1. Decision-making (DM) Orlovsky [53] defines DM as the act of a binary preference with a set of alternatives that are formulated and suggested to the decisionmaking person as their rational choices. According to Simon [63], DM involves choosing issues that require attention, finding adequate courses of action, and choosing an alternative as the final decision. Lendel [43] describes DM as a cognitive process that results in a final choice. It can be a selection of course of actions or opinions. The DM process starts with the reason for doing something in order to reach a decision. Beach [3] comments on DM research by highlighting the fact that decision theories have exclusively focused on choice – the selection of the best option or alternatives from a choice set containing two or more options – and that this is an incomplete view of the decisionmaking paradigm. From these definitions, we observe that the three main characteristics of DM theory to include: decision maker in the process of decision-making (cognitive process), alternatives (courses of action or opinion), and decision (final choice).

103

2.2. Decision making phases Simon [62] suggested that the decision-making process can be structured and ordered in three phases: intelligence, design, and choice. Huber and McDaniel [75] extended this model by adding two other phases: implementation and monitoring. The sequential model developed by Cooke and Slack [12] uses Simon's model to explain decision-making as a cyclical process that focuses around the problem. The problem solving process in their theory is not reflected in three distinct phases of the Simon model, but a continuous process of identifying the best alternatives and course of actions. The Mintzberg et al. [49] model follows linear design from Simon's rational decision-making process and reflects chaotic elements and incoherent phases of decision making. In this model, the decision maker comes with recognition of a problem or tangible request that requires an action, with the solution coming in a manner of different stages that do not necessarily follow a sequence. Unlike rational and sequential models, decision-making theories emerged into an anarchical problem-solving process that is driven by events. There is no sequence for decision phases and there is no established process to follow. There are chaotic and incoherent phases of decision making that build on need. In other words, this model argues for a free decision-making process that is more intuitive than rational [42]. The decision-making process driven by events is similar to Cohen et al.'s [10] garbage can model of decision choice. The four streams that interplay in Cohen's garbage can model are problems, solutions, participants, and choice opportunities. Cohen et al. [10] contradict the sequential modeling theory and state that DM streams are inconsistent and inter-correlate with each other like a vortex with no apparent structure or sequence. Langley et al.'s [42] convergence process is another approach in which a decision follows a trajectory and contrasts with sequential theory models. The decision comes in a more integrative way as the construction of issues. Rather than work backwards from the decision, this model works its way forward to a decision. The main weakness of these models is the loss of structure as well as poor description of what happens and how the decision maker reaches a particular idea. Sinclair and Ashkanasy [64] developed a model of integrated analytical and intuitive decision-making that supports two mechanisms of decision-making: first, Langley et al. [42] and Cohen et al. [10] view of the decision-making process as following an intuitive behaviour that is driven by events; and second, rationality in a decision-making process that is believed to be structured and logical towards problem solving. Complexity is explained through associations and metaphors of phases. To summarize, there are two main streams of models on decisionmaking: rational and anarchical, where the former emphasizes structure and sequence while the latter claim the lack of structure as well as sequence in the decision-making process. The establishment of decision-making theory goes back to the beginning of the 20th century. It is an obvious concern that with the involvement of technology and online interfaces, some of the models may not necessarily remain relevant to the current problem-solving environment. Another issue with the research conducted in the early last century lies in the methodology — a majority of these were conducted through interviews or organizational observations. This is a distant view on a real-time DM process. To understand the relationship between the decision maker and the process, the researcher needs to be involved in the actions of adopting phenomenological perspective. At the same time, to understand some global decisions, the researcher is required to zoom out and observe the overall phenomenon which might require an additional study into the history or expertise of the decision makers. The choice between either a closer look or a distant perspective depends on the purpose of the study, but what is obvious is the need to update the decision-making research view to incorporate

104

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

the reality of the changing world and high change rate of the environment.

2.3. Health Online Social Networks (HOSN) Even as late as 2004, social networks were largely ignored as a source of data, or reference for those seeking medical advice [8]. However, in more recent times, social networks have started to gain the attention of both the general public and those in the healthcare and medical sectors [2]. Moreover, health issues or questions are identified to be the top three most searched aspects on the Internet [44, 54]. According to a report released by the Pew Internet and American Life Project 2011, 59% of adults search online for health information. It is increasingly becoming common to seek for advice in Healthcare Online Social Networks (HOSN) [21,23], which may not necessarily provide the most appropriate information. A case in point is the refusal of some parents to have their child vaccinated as a result of (mis)information in some Online Social Network (OSN) about alleged severe side-effects ([50]). However it is apparent that properly governed and designed OSN/HOSN can play an important role in supporting different types of decision-making (DM) [42] as they provide their participants/stakeholders various forms of support [15] ranging from the instrumental to the emotional and informational. The evolution of OSN started from small friend-orientated collaborative networks, where pupils and students shared their profiles, ideas, interests and hobbies [38]. Applications such as Facebook, MySpace, Friendster, Bebo and Twitter have changed the world and its opinion on online interaction between people by removing geographical and cost barriers. A large number of users of face-to-face social networks have switched to online groups to establish connections with others who have had similar experiences, without the need to travel long distances to find the required support group [57]. We use the term ‘online social network site(s)’ to describe the phenomena of online interaction through social network communications. However, ‘social network site(s)’ also appears in public and academic discourse. Boyd and Ellison [5] define social networking sites (SNS) as Web-based services that provide the ability for stakeholders to (1) build a public or semi-public profile, (2) share the connection with other users within a bounded system, and (3) view and communicate with the list of connected users within a system. The nature of communication and connection can vary from site to site (p.211). SNS is more about already established relationships in the offline social world that are taken further to online communications with the goal of not necessarily to meet strangers, but mainly to support pre-existing relationships [5]. SNS is one of the first emergent definitions of computer-mediated communications, whereas OSN does not require any inter-connection from offline ties of networks and most often emphasizes the importance of strangers connected by OSN based on their common interests or purpose. For the present study, we are interested in the aspects of OSN and not SNS. We define ‘online social network sites’ as online public/semi-public/private services/sites/ platforms that facilitate the creation and reflection of social ties/networks/connections/relations among stakeholders/groups/organizations who share interests, activities, beliefs, dislikes, knowledge and/or values. Since health is one of the most important aspects in life, there is an instinct to share health information with others and get advice. According to Pew Research Online and the conference on Medicine 2.0, the Internet's popularity and one of the main purposes for using it is to share information, influence the healthcare sector and the connection of people with one another to improve their health [11,21]. The third most popular search topic online is indeed health, after sport and weather [20,21,56,60]. Online healthcare has the potential to transform traditional healthcare by allowing people to

share their advice, information and history with others through online interactions. Pew Internet and American Life Project has found “that six in ten U.S. adults gather health information online, one in five American Internet users have gone online to find others who might have health concerns similar to theirs. That percentage is even higher — one in four, among those living with chronic illness, those who are caring for a loved one, and those who have experienced a significant change in their physical health such as weight loss or gain, pregnancy, or quitting smoking” [21]. Many users of health online networks have “second degree” participation through a caregiver searching for health information on their behalf, and this is important for those who do not have access to the Internet [11,21,54]. The statistics on online healthcare information are a direct result of users' preference for HOSN rather than, or in addition to, other healthcare providers since there are no costs associated with travelling, time, and inconvenience. Macias et al. [44] observe that because of the rapidly increasing personal healthcare cost and less support from the government, stakeholders have become more responsible for maintaining their health. This has implications for the use of HOSN or other online health information providers more often because instead of consulting healthcare institutions, stakeholders are making decisions by themselves, where appropriate [44]. Therefore, it is important from a research perspective to understand how the Internet is used by people to gather information on health issues. To understand the concept of HOSN, it is important to introduce the primary users. In an offline environment they are the patients; in an online environment they are e-patients [19]. E-patients are those who seek online health advice, information, guidance for their own ailments and of their friends as well as family members [19]. Bichakjian et al. [4] explain the primary reason why users participate in HOSN: most people who search for medical information on specific health-orientated networks are often somehow associated with a defined problem; their family members, friends or they may be ill. Many e-patients appreciate the existence of online health support groups and the medical information and guidance provided creates the perception of being more complete and practical than that provided by healthcare institutions [18]. The reason e-patients transfer themselves into the online world for advice is that they can get free 24-hour support, seven days a week [9]. Furthermore, people with similar experience are connected with one another. Rodan et al. [57] observed e-patients on Heart.net OSN and found that some participants prefer night-time conversations. Most often, participants at this time express their innermost feelings with the assumption that no one else is watching, even though the online world is still active. E-patients, for the most part, act anonymously and are really protective of their identity [57]. Most sites provide the opportunity to register and participate in online forums, blogs or chat rooms in real-time and to send each other private messages. Many of the participants hide their identity due to the reason that their health questions or concerns can be too sensitive for public observation. It is important to note that HOSN are different in their structure, design and purpose compared to just healthcare information providers that operate online. HOSN usually offer real-life health stories, and provide emotional support, where people can express their thoughts and experience or have conversations with someone who went along a similar path, whereas healthcare websites provide information only and do not offer the communication environment [47]. Eysenbach et al. [18] observe that most researchers and health practitioners underestimate the benefits and overestimate the risks of available online health resources. Similar to Eysenbach et al. [18], Ferguson and Frydman [19] observe that the public get distracted by focusing on the negative aspects of OSN and HOSN and overlook its benefits. Although there are studies that illustrate and

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

measure the amount of inaccurate information in health networks, there are those that show how OSN are beneficial to users by providing emotional support and a comforting environment for communication and learning. The role of HOSN is not only to provide information but also to support stakeholders in their health DM [57]. Studies have observed that users who participate in HOSN found it more helpful than offline social groups, and oftentimes more cooperative than physicians [9]. 3. Research problems, requirements, and framework 3.1. Problems We identified three main issues in the integration of DM and HOSN that require attention. These can be classified as research and practical (Table 1). 3.2. Research problem-1 HOSNs have become a significant new phenomenon that attracts researchers and practitioners. DM with HOSN information requires a proper investigation in order to understand the effectiveness of HOSN to support decision-making and to provide information. 3.3. Research problem-2 Health and healthcare-oriented research have always grabbed the attention of the public. Through the involvement of technology, health-related research has become more innovative and successful. However, what remains a matter of serious concern is that there is no research-based model on how HOSN support the decisions of key health stakeholders such as patients, doctors, support groups, and the health care eco-system in its entirety. This presents a complex stream of issues that need to be explored in the area of HOSN and DM. 3.4. Practical problem Decision support with use of HOSN is a very delicate research topic. The actual decision support extent is hard to measure, and what supports one user will not necessarily be beneficial to others, especially in the case of HOSN. The most common problem in traditional literature on DM is that the models are set outside of the context of real decisions that should be exemplified in real decision situations by real people [42]. According to Lendel [43], DM is a chaotic process that cannot be simply plugged into a model of action; good decision-making is not a predictable pattern, but can be nasty, narrow, scattered or fuzzy. While some of these issues are embedded with the HOSN decision makers/users, others are orientated towards HOSN design and the ability to meet user expectations. However, the issue with research on HOSN is the lack of evidence on how HOSN influences the DM process of stakeholders. 3.5. Requirements To overcome the previously stated problems, we propose a set of generic and specific requirements that should be addressed and further employed. Table 2 provides an overview of relationships among the problems and requirements of this study. Two types of requirements Table 1 Overview of DM in HOSN problems. Research problem-1

Research problem-2

Practical problem

Dynamics of DM process and use of HOSN for decision support.

How does HOSN support the key health stakeholders' decisions?

How does HOSN influence stakeholders' DM process?

105

are proposed for the three recognized problems. Generic requirements present an overview of the solutions; specific requirements are associated with possible solutions, actions, and potential artefacts. These problems present a complicated structure of research and practical gaps that have been over-looked and cannot be solved simply by a provided single resolution. The theoretical problems on lack of research lead to the practical matter of lack of evidence on how HOSN supports the DM process. Review of existing literature indicates that there is a vast amount of research on social networks (e.g., [7,74]), decision-making (e.g., [58]), and healthcare (e.g., [68]). However, research is needed on (1) how online social networks support decision-making and (2) how HOSN support the decisions of key health stakeholders such as patients, doctors, support groups, and the health ecosystem as a whole. The nature of these questions and the domain of the research demand a research approach that begins with exploration and progresses towards explanation. Given its natural fit, we use netnography for this purpose.

4. The netnography research process Netnography is part of the ethnographic research family. Kozinets developed netnography as an online marketing-research technique to study consumer insights [36]. It is the study of fieldwork, distinctive meanings, practices and artefacts of human behaviour, societies and cultures [46]. Netnography provides rich insights into the human society [35] and is different from ethnography in that it is conducted through online communities with computer-mediated communication, whereas ethnography is purely an anthropological approach where personal engagement is necessary for the study of a particular fieldwork or social setting [31]. Ethnographers immerse themselves in the life of people they study [52]. With the appearance of online communities in the research field, this methodology has emerged as ‘online ethnography’, sometimes referred to as cyber ethnography [22,71] or netnography [37]. Based on research objectives that focus on online social networks, decision-making process and decision makers, our goal is to capture decision-making experiences of online users. Netnography assists in understanding the behaviour of online users and their experience in the DM process. We use netnography to answer questions that include, “How does HOSN support the DM process?” or “How can it support the process, and how can it be evaluated and investigated?” With netnography in online communities, we can answer subsequent research questions on identifying those phases of the DM process that are supported by the use of OSN. The use of netnography makes particular sense for attempts to analyse communities where access to information is strictly restricted or difficult. Health topics that involve a DM process around cancer, depression, drug addiction and pregnancy involve a great deal of privately held information. Netnography helps extract the voice of a specific population, where interviews or other research methods are not capable of providing insights because of reasons such as ethical norms, participants' willingness, research cost, time and geographical constraints. Since the goal is to study the DM process of online health participants, it is sensitive to study due to ethical standards as well as the highest privacy standards of health information. We chose observational netnography since this research study is associated with aspects of health, and health-related topics fall into the category of sensitive research topics. If the research information can be accessed without an actual participant and/or the informants are unwilling to communicate with the researcher, then ‘passive” netnography is a suitable methodology to study sensitive topics [41]. We followed Kozinets' [37] guidance on conducting netnography, with appropriate modifications as necessary. Kozinets [37] identified five overlapping steps of netnographic research (Table 3).

106

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120 Table 2 Research problems, generic and specific requirements.

Problems

Generic requirements Study the area of the research

1. DM process and use of HOSN for decision support.

Explore and understand the various types of HOSN

Understand the different stakeholders & users of HOSN and their needs/requirements.

2. How does HOSN support the decisions of key health stakeholders

Understand the elements & factors that affect DM in HOSN Extract real life examples as an evidence of how HOSN supports the DM Process

We closely follow methodological steps including entrée, data collection, analysis and interpretation as per Kozinets' guidance. For ethical consideration that includes member checks, we reconsidered and modified the rules as necessary and appropriate. The reason for modification of this particular step can be explained by the fact that HOSN involves a sensitive research topic. Therefore, by simplifying the ethical guidance this study could get a broader knowledge of networks and gain full advantage of the online method.

4.1. Planning and Entrée The aforementioned research objectives and questions guided the choice of the appropriate online communities. Kozinets [37] provide a list of six criteria that help choose a winning community of study.

Table 3 The netnography research process. Adapted from Kozinets [37]. Step 1: Planning and entrée Step 2: Data collection Step 3: Data analysis Step 4: Conducting ethical netnography Step 5: Representation and evaluation

Observe DM process by use of HOSN Investigate DM process by use of HOSN

Identify different phases of DM process that are supported by HOSN.

Explore the process of DM that people follow by use of HOSN

Explore and understand how HOSN supports DM and in particular various phases of DM and various styles of decision-making and decision maker participation style 3. How does HOSN influence the DM process of stakeholders.

Specific requirements

Identification of research questions, online communities, HOSN of interests and types of network USERS Data filtering process, data collection challenges Application of the analysis process Not applicable The interpretation process: phases of the HOSN–DM process

Investigate the environment of HOSN, and study the support HOSN features that can fulfil the purpose of the DM.

Propose DM process model by use of HOSN

Build a generic framework of well governed multistakeholder supportive HOSN that addresses the identified research problems to guide a design.

These criteria of choice are relevant, active, interactive, substantial, heterogeneous, and data rich. The first step in choosing a community of interest is not problematic given the multitude of available HOSN. The problem is to find relevant communities with discussion of the DM process. The community of interest needs to be active; it should have regular and recent high posting traffic. The communities must have communication flow among participants. Interactivity involves participation of different stakeholders in the conversations. Substantial communities in health networks are hard to find. There are many networks with flashy features and a diverse audience but not many have the energetic feel of quality conversations, which involve DM or problem solving situations. Communities need to have participants from various age groups, locations and sexual orientation. In healthcare DM, it is desirable to find communities where additional related information are provided, where posts have a full description of the story, it is also important to have relevant advice and qualified expert answers from different stakeholders. Entrée involves the screening of online communities that are most appropriate to the research objective as well as gaining as much information as possible to add value to the research. It is essential to find online social networks where the experts, caregivers, patients and just random users interact with one another, not necessarily simultaneously or at one network. The typologies of online communities were not barriers for the research, as health communication could find existence in different types of forums, bulletin boards, independent Web pages (e.g., hospital sites) as long as the conversation flow is consistent. Each of these types of online social networks provided different insights

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

to the study that are valuable for the analysis part. Kozinets [37] recommends that conversation and topics should be popular with much interaction and visitors. We decided to choose the top 11 most popular health topics discussed on the Internet: Asthma, Alcohol and drug addiction, Bulimia, Cancer, Depression, General Health problems/ medicine, HIV/AIDS, Plastic Surgery, Pregnancy, Scoliosis Problem, and Weight Loss/Appearance issues. These are the most popular and engaging topics of health networks (HSAfinder.com [32], Sutterhealth.org [66]). We chose topics based on observed conversations and existence of DM process phases. We chose topics such as cancer and AIDS/HIV primarily based on their popularity and importance to society. Even these networks/conversations could not provide rich insights into the already existing theory of the DM process; they could indicate the existence of emerging steps/phases of DM that can only be associated with HOSN. One of the research objectives enquires about HOSN users, their typology and their interactions with one another. Researchers have classified online participants similar to that of Kozinets typology — Tourists, Minglers, Devotees, and Insiders. With our requirement for typology of HOSN users who participate in the networks and the ones who are seeking support in the DM process, we classified the participants as Advisers, Seekers, and Observers. Observers are less associated with community life, are searching for the appropriate information to support their decision or are there to just find some clues/interest to questions or answers. Seekers do not always have strong ties with an associated group. They are brave to ask questions and find support, and are interested in activities around the associated group and advice provided by advisers. Advisers have strong associated ties with a group, a high rate of participation, and devote a strong interest to the group. Advisers are those who provide support to seekers in order to solve a problem. While advisers can support decision makers, they also have the potential to mislead them. Advisers and Seekers are the most important participants who bring the content, information and interest into the group. Our netnography study is focused primarily on these two types of participants.

107

the observation ideas and comments were recorded and synthesised in the analysis section. The direct quotes that are used in the analysis section were not corrected for typographical errors and have been copied directly from the discussion forums. We followed the five W's analysis tool which attempts to filter audience conversations and create the general idea behind the context of the conversations, understand if it requires attention and if it follows any phases of DM process. The five Ws approach can discover information necessary for the analysis of the audience's needs [30].

What were the actions and what happened as a result of the actions? Who performed the actions in the story (or who experienced the results)? Where did the actions occur? When did the actions occur? Why did the actions occur?

Five W questions indicated that the problem exists, background information of the problem is provided and that the seeker is looking for an answer in order to solve a problem/make decision, or a seeker is waiting for an adviser who can help by providing personal experience and an already implemented solution to similar concerns. Most of the participants share their DM process experience under the specific headings of the conversations. In this data collection process, we paid high attention to the posts that had a subject heading that included one of the five W question words: When should I get a nose job? I'm so depressed — What should I do? What kind of medicine should I take? I always have a headache what kind of medicine should I take? What is better for fat loss protein or ceratine?

4.2. Data collection The second step in the netnographic research framework involves the data collection process. We adopted two steps of the data collection process as recommended by Kozinets [37]. We obtained written communications among different stakeholders that occurred in online communities. We also prepared self-authorised field notes in which

With the use of Ws, we observed that the conversations that started with the question mark are most often associated with a DM problem in which a seeker is trying to find an answer, get support or seek advice. Examples of these decision-making conversations are:

Fig. 1. Decision making posts' review process.

108

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120 Table 4 Data analysis: Step 2 — formation of new reviewed categories.

Fig. 2. Qualitative data analytic process. Adapted from Miles and Huberman [48].

How can these therapies treat cancer!? herbal medicines for cancer, yoga for cancer, acupuncture for cancer? Do I have Cancer? Getting a nose job, is it worth it? Are my implants making me sick?

After reading and screening hundreds of message boards, another pattern that indicated the existence of a DM process includes posts that ask for help. Threads that have the word ‘help’ in their heading usually attract most of the replies, advice and alternative models of how to solve a problem. Examples of ‘help’ orientated threads are:

Pics of me: Help me choose procedures as I'm butt ugly! Addicted to Adderall and Alcohol!!!! NEED HELP!!! I need to quit drinking… I need help… Please help me? I CANT STOP BINGING… NEED URGENT HELP These trends of DM conversation emerged from the process of screening and filtering message posts. We used these emerging patterns to find more conversations that are associated with the DM process. The second part of the process is identification of health conversations and categorisation. Message boards were found by examining top-level headings with the highest rating and highest number of participants or by a search for words related to a chosen health topic. We searched for applicable posts by use of search engines like Google, Yahoo, and Yahoo Groups. We chose 29 websites with conversations relative to the research questions on the DM process, phase of DM, and representations of different HOSN stakeholders in the discussion about health matters. The communication language in these websites is English. Furthermore,

S — Seeker

A — Adviser

IS — intelligence provided by seeker SM — seeker provides a model of a problem instead of Design (D), the category was renamed as Model (M) SC — choice taken by the seeker SI — seeker provides an implementation of past/current or future decisions MS — monitoring result provided by seeker of implemented decision

IA — intelligence provided by adviser AM — adviser provides design model to seeker that includes alternatives and options CA — choice provided by adviser AI — adviser provides an implementation for the decisions MA — monitoring result provided by adviser

most of the websites were hosted in the US although the users themselves were distributed all over the world. The data collection process started in March and was partially completed by the beginning of August, to secure sufficient volume of conversations. During this period 51 full conversations were downloaded from 29 websites, filling 343 pages with an average of 878 words per post and an average of 17 messages per post. Statistically, one website could provide more than one conversation for analysis. Each health category could appear on a maximum of four different websites. The effects of bandwagon and confirmation biases of information processing were alleviated by choosing three to four different topics under each health category. Through this analysis the conversations were not on a single subject and were not conducted by the same participants; therefore the decision-making process is not repetitive, and different sets of participants represent the voice of different demographics. We printed, coded and categorised all contributions for further analysis and interpretation. We classified all 51 posts by the subheading of the conversation. We collected and separated conversations by subject of interest, phases of decision making and conversation headings. We assigned a specific code for each post that indicates its subject and number of posts in this subject group. Each post has statistical information regarding the number of conversations, number of words and number of participants. All 51 posts with a total number of 734 messages were chosen through the main research criteria which required that the posts have conversations regarding the DM or problem solving process. It is also important to see how different DM phases are supported by HOSN. Fig. 1 shows the review process of collected posts, where all the messages that contain a full description of the DM process were reviewed and then the posts that were lacking detailed information or have been off topic were discarded. The data collection process is a big challenge, especially when using the Internet as the main data source [27]. There is always a danger of getting lost with an excessive amount of data and field notes [37,76].

Fig. 3. Adapted data analysis process.

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120 Table 5 Data analysis: Step 3 — final revision of categories. New steps of DM process New categories of DM phases

EX — exit, E — entry, SE — seeker entry IS — intelligence by seeker SM — seeker model SC — seeker choice SI — seeker implementation MS — monitoring seeker

ES — emotional support AE — adviser entry IA — intelligence by adviser AM — adviser model AMA — adviser model based on his experience AO — adviser option AA — adviser alternative CA — choice by adviser AI — adviser implementation MA — monitoring by adviser

One of the data collection challenges was the search procedure for the appropriate conversations that relate to DM or problem solving at least from the pool of millions of health-related conversations. We addressed this by filtering online posts by headings that have a five W question or one that asked for help or started with a question. Another challenge was the vast number of HOSN that generate a large number of posts every day. The ‘dangerous’ thought is that there is always more data out there that can bring more insights. The solution to this is in research planning, closely following the research process, absorption of research objectives and questions only, restriction of the health sector to 11 topics and the identified approximate number of conversations per topic. 4.3. Data analysis Data analysis process in a grounded theory approach has a generic sequence of common qualitative steps. Kozinets [37] follows the procedure described by Milles and Huberman [48] shown in Fig. 2. While following the Kozinets [37] proposed sequence, we have adapted it to suit our requirements. We first identified a theoretical framework, which is supported by Simon's [62] DM model, hypotheses were then proposed that tried to identify how HOSN support the DM process, and only then could we proceed with the coding, abstracting and conceptualising that has been exercised in this study. We followed a traditional fashion of qualitative methodology in this study because

109

the research initially was motivated by Simon's [62] theory and how it applies in the HOSN world. Fig. 3 illustrates how we followed Kozinets' [37] steps in a different sequence. In this analysis from a higher level of abstraction, we went back to a lower level — that brought more details and information. Our data analysis process was found to be more disordered than structural. Spiggle [65] argues with grounded theory scholars by explaining that the process of interpretation and analysis can be chaotic and should not necessarily follow a particular structure. Therefore the data analysis in netnography cannot be viewed as an assortment of procedures and methodological steps. The grounded theory guidelines were not ignored and served more of a foundational and manual purpose on how to analyse data. The actual analysis began when the collected data started to fall into the particular categories and the pattern of findings could be identified. These categories came from the research theory and research objective. The initial categories were taken from Simon's [62] DM model, where he identified that a rational human mind goes through five phases of a DM process, those of Intelligence, Design, Choice, Implementation and Monitoring. The data analysis process went through four stages (Fig. 3). First, high level of analysis — the general and theoretical understanding of the collected data and identification of keywords took place. Second, identification of main categories and research concepts — emergent concepts of decision-making phases were evaluated and grouped under categories. The third stage of analysis involves deeper exploration of data and general findings from stages 1 and 2. In revising original categories, we looked for additional evidence and facts that could strengthen theory-building. At the final stage of the data analysis process, we revised the theory, categories, associated dimensions and codes, with the purpose of discovering emergent concepts that may have been overlooked in earlier analysis steps. High level of analysis — we coded the data into discrete parts of five phases of the DM process. Texts were coded according to Glaser and Strauss' [26] method where constant comparative analysis was implemented by reading the conversation several times and grouping them into specific categories. We formed specific categories by identifying keywords that are associated with each phase of decision-making

Fig. 4. Structure of the DM conversations.

110

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

Table 6 Coded content. Number of total topics Total number of sites Total number of messages Total number of posts Total number of words Total number of seekers Total number of advisers Total number of intelligence phase provided by seekers (IS) Total number of intelligence phase provided by adviser (IA) Total number of design phase provided by seekers (SM) Total number of design phase provided by adviser (AM) Total number of choice phase provided by seekers (SC) Total number of choice phase provided by advisers (CA) Total number of implementation phase provided by seekers (SI) Total number of implementation phase provided by advisers (AI) Total number of monitoring phase provided by seekers (MS) Total number of monitoring phase provided by advisers (MA)

11 29 734 51 72,003 69 550 102 328 62 408 39 187 44 132 22 103

(Intelligence (I), Design (D), Choice (C), Implementation (IM), Monitoring (M)). The second stage of the analysis process is about revising stage 1 and conceptualising the new categories and new codes. Table 4 shows categories of five phases of the decision-making process initiated by seeker and adviser. We identified that the conversation involved two groups of participants, seekers and advisers, and therefore the phases of decisionmaking is grouped by participant. Ten categories emerged from the theoretical framework. They were revised several times and responses were re-checked and grouped to revised categories according to associated criteria. This procedure resulted in the new themes that have been added to categories, and a new process that outlines how HOSN support the decision-making process. It also helped to understand how the DM process takes place in the mind of online participants. In the third layer of analysis, with additional exploration of data, literature and pre-findings results, additional phases as well as new categories of the DM process were established for further analysis and coding. These were based on existing phases of Simon's [62] DM model, but with the addition of the use of HOSN. The new steps were identified by observation and comparison of multiple conversations. Table 5 shows how we expand the second analysis stage by adding new steps and categories as observed in an online decision making environment. The difference between real and online worlds is how people present and describe their future decisions or past DM processes. Oftentimes, an adviser provides a model of decision (design phase) based on previous experience of decisions, knowledge and insights. Adviser model can include options or alternatives provided by adviser. Alternatives are identified to be different from options. Alternatives often contradict with seeker choices or possible solutions to the expressed problem statement. The adviser's personal experience expands the categories of the DM process. Through revising the conversations we observed that adviser and seeker follow the real life conversation situation. By entering the community they familiarise themselves with it, and by exiting the community they repeat the same procedure as in real life. These new steps are assigned to be an addition to the already developed structure of phases of DM process. An example of entry and exit is presented below from the cancer network message.

Entry: Dr. X, You made claims about drinking and massaging with urine that I would like you to substantiate: Exit: With the very best!

The background information provided by a seeker or adviser is identified as a DM process entry. HOSN users often provide their background

information that is not necessarily related to the five phases of the DM process. The new DM step that was found to be substantial and present in most of the conversations in HOSN is ‘emotional support’. There are always supportive and wishful thoughts that go to seeker from adviser or seeker is thankful for the support provided by the community. It is not always the case that a seeker in real life can get so much emotional help from outsiders compared to online conversations ([18,67,72,77]). Theories on DM process do not indicate the fact of how emotional support influences a decision. Our study emphasizes the importance of the fact that more support exist in OSN compared to the outside world. The following HOSN examples show people express their emotional support and willingness to help.

I thank God for putting in my heart to get online, so I could read your story and now I could pray for you.. He's gonna help you get out of this mess.... Don't worry about tomorrow…GOD is already there! Your friend needs to make sure that her daughter is safe right now, no matter what. I will keep her in my prayers tonight. Please feel free to pm me anytime. My heart goes out to you. My thoughts and prayers are with you, and please try to stay strong

For the final stage of the analysis process, we start with the final revision of collected data and emerged categories through the first three stages. We re-read and recorded 51 downloaded posts according to the new categories that were described previously. The reason for repeating the revision more than twice was to go back to the roots with a fresh mind-set and deeper understanding of online communities, health stakeholders, DM theory, DM phases and its emerging patterns. Also, the final stage revealed that the DM process has a chaotic behaviour in the online environment and the phases are not necessarily connected in order for a decision to be made. Visualisation helped to show the most frequent phases and construction of conversation for decision making that exists in health online social networks. Appendices A, B, C, and D present the data analysis spreadsheets and diagrams on the structure of the DM conversations from cancer, weight loss, general health and plastic surgery networks; the analysis illustrates the phases of the DM and assigned categories with new emergent steps. Fig. 4 illustrates how the phases of DM process are interconnected in the online environment that makes the DM process follow an anarchical structure. The DM process phases are visible, but the sequence in which conversations move between them is unstructured and appear random. Here, the (blue) rectangles shows the DM process (known Simon's phases); the (grey) ellipses represent the identified additional aspects of the HOSN-DM process that go beyond the DM phases and include Entrée, Background Information, Emotional Support, Exit — we include this to show that most of conversations have these aspects, which is not necessarily mentioned in the DM and OSN literature; the diamonds represent the points at which the conversation started (the moment when adviser or seeker steps in into the conversation), and if there's a number, it represents the number of participant (if number is repeated, it means that the participants walks in and out to conversation several times); the conversation arrows from each diamond show the pattern and sequence of the conversations (i.e., I–D or I–C–I–M or I–D–I). Sometimes the DM process (conversation) can end at the Intelligence or Design phase — the sequence is not structured, it is anarchical as it depends on the questions/problems of the conversation that are posted in the OSN and behavioural patterns of seeker and adviser. The relative position of different components in Fig. 4 has significance. For example, ‘Entry’ and ‘Background Information’ normally occur at the beginning of the decision making phases just prior to the ‘Intelligence’ phase. Hence, we have positioned them before the

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

‘Intelligence’ phase. ‘Emotional Support’ usually occurs in conjunction with ‘Implementation’, post-implementation, and ‘Monitoring’ phases. Hence, we have positioned it close to them. ‘Exit’ usually occurs after the ‘Choice’ phase where many decision makers exit the conversation. We followed the netnography and grounded theory guidelines in data analysis, where some steps of grounded theory were expanded and some were omitted due to the research outline. We took a multianalysis approach, where grounded theory brought the theory into analysis, but the content, conversation and discourse analysis added imperative details, quantitative view and richness to the entire research process and findings [37]. We use statistical content analysis in this netnographic study to identify the weightings of the five phases of the DM process, and to identify and understand the source of content. Table 6 shows the content coded and analysis of data in this research. As can be seen, advisers dominate each phase of the DM process. Statistically, on each seeker concern posted on the wall of HOSN there were approximately 7.91 replies. That indicates that there are more advisers in HOSN than seekers, and it can be assumed that advisers are willing to share their health experience with others or provide information regarding health concerns. As for DM phases, the design phases of advisers have the highest score, meaning that advisers propose the models for decision-making that consists of alternatives and options that can accommodate the seeker query. The actual number of implementation phases of advisers shows how many times advisers shared their decision experience with the HOSN public. There is a good pattern of choice phase from the seeker prospect which signifies that 39 times the seeker mentions the fact of making a choice in online conversations. Some seekers have returned to the network to admit the fact that the choice has been made. These figures show a numerical inequality between the number of participants and their support for the DM process. Thus, the conversations on HOSN do not necessarily reach every phase of the DM process. This entails the question of the behaviour of participants, their participation style and DM style. Conversational analysis (CA) is used in this study to describe the orderliness, structure and sequential pattern of participant's interaction in HOSN. As previously described, the conversations were broken down into phases of DM process, and new steps and categories were identified during the analysis process. New steps such as entry, exit and background information showed that HOSN conversations were following real-life structure, where participants introduce themselves by providing background information and exit the conversation when required. However, the emotional support step showed that HOSN conversations are unique in their expressions of support from participants. Table 6 presents the statistical breakdown of conversations that indicates that seekers are not active participants as in real life, most of the time not following the phases of the DM process, but just questioning or posting queries on HOSN walls. Moreover, seekers have a pattern of disappearing after they post a question. The conversational analysis also revealed the phenomenon of adviser experience: the adviser tended to follow the sequence of phases of the DM process when describing decision experience. The conversational analysis helps to answer the research question on the phases of decision-making that are supported by the use of HOSN. Nearly all phases of the DM process have been attended by HOSN users at least once from each post. The discourse analysis helped us understand the underlying mood of conversations, the context behind the texts, participants' thoughts that were not expressed but can be understood. While discourse analysis enabled us to analyse transparency on HOSN, the major problem faced was the transparency of phases like choice, implementation and monitoring. Discourse analysis helped identify the conversations that are more advertisement than advice, the fact the doctors at no time gave clear option/advice for a seeker question/query, and the kind and supportive language used by users to not embarrass or hurt each other. Discourse analysis also helped analyse pseudonyms that users

111

use in HOSN, the chosen names were usually associated with the diagnosis they have, problems, their hopes, or current mood. 4.4. Findings and discussion 4.4.1. Phases of the HOSN-DM process Our data analysis revealed 15 categories of five DM phases that can be identified in the DM conversations through use of HOSN. We discuss these in the following paragraphs. 4.4.2. Entry/exit by adviser and seeker (EA, ES) Entry by seeker (ES) is more informative than that of adviser. A seeker usually enters the network by introducing himself/herself to the community, and providing some background information about himself/herself. The seeker then provides details on the decision or problem that needs to be addressed. If an adviser uses the exit technique (EA) of communication, it signifies that the conversation is over and the adviser provided the information or at least assisted with a solution that was required by the seeker. An example from the cancer network, where a seeker indicates her age and gender as background information and her wishes in life:

Seeker Hi all I'm a 28 year old female. And just been diagnosed with cervical cancer. They say that only 25% of women my age get diagnosed with cervical cancer. Will like to talk to women with the same thing. I'm so scared..... I have 2 kids one 7 and the other 2… I hate to think that i might never see them grow up. And what is worse i always hoped that i will have a boy. Now that is out of the question.

4.4.3. Intelligence by adviser and seeker (IA, IS) The intelligence phase is a discussion between seeker and adviser about recognition and classification of a problem or decision. In a health case it is a diagnosis or health issue information. The example below shows the intelligence phase from seeker (IS), who provides information about a diagnosis.

23 y.o. w/ recent chest pains, PLEASE respond Seeker 1 Hello everyone, this is my first post in this forum and I am very happy to have found this place for support. To give a little background on myself: I'm a 23 year old male who has had I would say mild kyphosis for quite a while. I always just thought it was a normal thing to have to constantly pop my back to make it feel better/ straighter until a couple years ago. My physician didn't even diagnose it without me saying could you look closer at my back it feels curved and I'm constantly having to pop it. Then after he looked at it closer he said oh well you're right, it looks like you have a mild case of kyphosis. (hurrah for doctors, right?)…I am worried… Could this kyphosis permanently have damaged my heart? It's weird, frequently my chest feels a little tight, I then pop my back, and then it feels completely normal again. Please give advice, hope, honesty, anything. I worry…

4.4.4. Model by adviser and seeker (AM, SM, AA, AO, EAM) The design phase in the DM process is about the formulation of a model of alternatives, options, course of actions in response to the situation described in the intelligence phase. AM is a model for a decision or solution that is provided by the adviser, and describes the process of problem solving with all possible outcomes. It is not a plan of implementation of a chosen solution; it is a description of possible ways on how to come to the choice phase. SM — is a model of the problem or solution that is described by a seeker. From the previous example, the

112

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

Table 7 Support of DM process in HOSN. DM phases

HOSN (netnography)

Intelligence Design Choice Implementation Monitoring

Supported Supported Supported through adviser sharing previous experience Supported through adviser sharing previous experience Weak support

Adviser Implementation (AI) category can support the seeker implementation process by describing the decision implementation. Since the implementation phase is not always transparent in HOSN, the use of discourse and conversational analysis helped to identify some patterns of how implementation is textually presented in online conversations. In the example below, the adviser admits to being an alcoholic and implementing visits to AA as a solution to quit. The adviser also mentions where to find AA contact numbers and the cost of service.

seeker model is presented by his situation of the physician not realizing his cause of problem: “…My physician didn't even diagnose it without me saying could you look closer at my back it feels curved....” Adviser Alternative (AA) is when the adviser is proposing an opposite solution to the seeker's choice. Adviser Option (AO) — is usually just an option that the adviser proposes to the seeker without possible outcomes or detailed explanation.

Seeker: I'm a functional alcoholic and have been for some time; my hubby can't take it anymore because he can't function and I don't want to lose him. I love him so much. I have to quit drinking and I need help. Please, someone.. give me some insight on what it takes… Thank you! Adviser: I went to AA many years ago and it is working for me. They are online and in the phone book. Look them up, it is free. Blessings

Adviser Hi, my name is ——— and I also have kyphosis. Sorry to hear that you are experiencing pain! I know exactly what you are talking about with the needing to “pop” your back. I myself do not experience chest pain, but my chiropractor told me that if my curve gets worse I may. From what he said, I guess there is a nerve in your mid-back that wraps around to the front and can cause chest pain if it is pinched by your curve. I would think that this is what is causing your pain, but as I am no doctor if you are worried about it I would go to your physician and see what they have to say. I hope this helps and good luck.

4.4.7. Monitoring (MS, MA) Past choices are assessed by the seeker or adviser by sharing their own experience about a past decision. After the seeker has decided and implemented the decision, the next phase is post-analysis of decision and monitoring of the outcome. This phase is also not transparent for the observer; rarely does the seeker come back to the conversation field to talk about the monitoring phase. As for adviser monitoring phase (MA), the example from alcoholic addiction HOSN shows the monitoring activities of adviser by the phrase “… it is working for me”.

EAM — model by adviser based on his experience. An example from the same post shows the adviser response where adviser explains his experience with back pain, along with explanation that his doctor provided and then adviser provides an option to a seeker “… I would go to your physician and see what they have to say…”. 4.4.5. Choice (CA, CS) The choice made by the seeker or adviser in most of the cases from the undertaken observational research is not transparent in the text conversations, but it can be identified from context by use of discourse and conversational analysis. Choice by Seeker (CS) can be confirmed only if the seeker comes back to the community and admits that the choice has been made based on conversations that were generated on the network(s). Choice by Adviser (CA) happens when the adviser proposes a solution to a problem or plan of implementation or provides a choice that has already been used by other stakeholders, institutions or organizations. Adviser also prefers to provide the problem solving choice from previous real-life examples from similar decision-making situations. An example below shows how an adviser provides a choice of cutting down on asthmatic medicine since it helped him relieve his concerns.

Seeker What medicines can I take to help relieve my symptoms of asthma? Adviser .....I've had asthma all my life, and the best relief I ever got was to cut back on my medications. I am at the point now where I only keep two on hand, predisone and Benadryl….

4.4.6. Implementation (AI, SI) Seeker Implementation (SI) — not always feasible and exists only if a seeker returns to the community and confirms the implemented choice in real life. It can be explained by the fact that implementation is an actual ‘doing’ process that does not involve textual presentation (texts are our thoughts) and that actions are more about physical processes.

Adviser: I went to AA many years ago and it is working for me. The discourse and conversational analysis were applied to the nontransparent phases of DM process. It helped to illuminate the weakness of not identifying DM phases by analysing the subtext of written text. 4.4.8. Nature of the HOSN-DM process One of the major findings of this netnographic study is that the decision makers follow a chaotic DM process and do not follow the rational DM processes developed by Simon [62]. By observing online communities and conversations, it could be seen that decision makers who use HOSN for problem-solving are not necessarily rational and most of the time are irrational stakeholders who follow an anarchical DM process. However, the rationality does not affect the support function of HOSN, which maintains support for most of the phases of DM process. The support function depends on the type of network, type of stakeholders who are engaged in decision conversations and types of issues about which seekers enquire. The DM conversation usually starts with a seeker expressing a concern regarding health issues (intelligence phase). An adviser in response provides a model to a seeker to solve a problem (design phase), and the model is then explained in detail, which is the intelligence step in the design phase. After the description of the model, the adviser can leave the network or provide personal experience on a similar decision, starting discussion from the implementation phase by admitting that option/alternative has been implemented or by providing results of an implemented decision (monitoring phase). Therefore the choice, implementation and monitoring phases from an adviser experience do not have any limitations and are interconnected with others by acting sequentially as described by Langley et al. [42]. Observation of HOSN shows the existence of emotional support in the DM process; an emotional support step is found in any stage of the conversation. There seems to be no pattern on how stakeholders build conversations online; these represent their cognitive process. The full support of the DM process is found in those conversations

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

that include description of models, options, alternatives (design phase), and adviser experience presented by choice, implementation, and monitoring. Adviser experience is observed as an essential component of the HOSN DM process. The only sequence that could be identified in the DM conversation is that the monitoring phase comes mostly after the implementation phase. But intelligence, design and choice don't have sequence rules. Even though real-life conversation starts with intelligence, in OSN advisers can provide models or choices and sometimes even implementations from past experience without any introduction or background. This is really a phenomenal aspect of conversation through computer-mediated communications. Another point observed through netnography is that actual experts (doctors, medical representatives) are not providing a concrete choice phase. They advise on further action, with the most common advice being — ‘go and see your doctor.’ They can also answer a seeker's questions in an informative way, but do not take responsibility for a choice phase. Therefore, most of the professionals in HOSN provide the intelligence phase as a source of information and the model phase as a source of a feasible plan on how to come to the decision or solve a problem. The example below shows an expert advice on the seeker query regarding the healing procedure after rhinoplasty: Adviser “…at this point it would be best for you to speak with your surgeon as he or she knows the extent of your surgery and exactly what was performed. Thank you very much for the question…” Based on the analysis and discussion above, we can confirm the answer to the main research question in the affirmative that HOSN does provide support for the DM process. Netnography helped us understand the main motives and reasons as to why a decision maker chooses online over real-life networks. Empirical findings verify that consumers/ stakeholders use Internet message boards in order to exchange information/provide support/give support/make decision/advise. 5. Discussion and conclusion The primary purpose of this study is to understand the decisionmaking process when HOSN is used as a support tool for DM and to identify the phases that are used by HOSN users for DM. Based on our analysis through netnography on HOSN, it is evident that online conversations support most of the phases of the decision-making process. Our results indicate that the phases in online conversation do not follow Simon's [62] sequence of a rational decision-making process. However, Simon's study mainly concentrated on analysing the behaviour of rational decision makers in the DM process. Our netnography results revealed that decision makers' conversations and textual thoughts do not follow the rational (sequential) DM process. The main thoughts, discussions and considerations that took place in this study were bounded around the subject of the DM process and how it can be supported by HOSN. The main finding of this research and its contribution to the theory and practice of DM and HOSN can be summarized as follows: Rational models of decision-making emphasize structure and sequence while anarchical models of decision-making imply that there is no structure and sequence in many real-world decision making contexts. However our results challenge both these models by suggesting that decision making in HOSN exhibit structure but not sequence. 5.1. Overall research findings Findings of this research study suggest that certain DM processes observed in HOSN networks, from an overall perspective, are in some way related to the well-known model of anarchical DM process driven

113

by events developed by Langley et al. [42]. The impulsive phases of the DM process are assembled accordingly to the needs of the decision-maker. Also, the observation validates the pattern of the DM process of interplaying with four streams (problems, participants, solutions and choice opportunities) [10]. A decision is generated by various opportunities, alternatives, associated problems and people. Discussions in HOSN environments involve stakeholders as advisers or seekers, models of choice, with alternatives and possible options that can be advised by people or provided from their experience. Observed HOSN DM process was found to have no structure and follows an anarchic behaviour. The observed HOSN DM process also displays characteristics similar to the Mintzberg et al. [49] model of the DM process as an iterative sequence. Through synthesising the findings from netnography, it emerged that HOSN supports decision makers in the most well accepted phases of the DM process. Based on our analysis, we observe that HOSN provides support for intelligence (information search), design (model of alternatives and options), and choice phases. This finding is aligned with the original Simon [62] model of the DM process. Table 7 shows the comparison of netnography results with respect to the research question: “What are the phases of DM process that are supported by use of HOSN?” An explanation for the observation of only partial support for phases such as choice, implementation and monitoring can be given from the theory of the cognitive process of the decision maker. In the realworld, decisions are made unconsciously in the mind of the decision maker. From the perspective of the decision maker there is no intention of admitting the decision choice, implementation and monitoring by the textual impressions in HOSN. Implementation and monitoring found only partial support from the observation of HOSN because these phases are not always feasible; they present an objective process that is not apparent in HOSN. Using discourse and conversational analysis for data interpretation from netnography, we observe that choice, implementation and monitoring phases are present but only as a subtext of the conversations; or they can be feasible only if extracted from the decision maker's experience. From the discussion above, it is apparent that HOSN is used as a support tool that helps find relevant information, understand alternatives, options, choices and consequences. It is also useful for observing and sharing the DM process experience, identifying necessary resources for implementation and evaluating outcomes from taken decisions. Our study discovered new steps of the DM process in HOSN not apparent in existing literature. First is the use of ‘experience of others’. Based on the observation of HOSN, it was apparent that most of the advice, options and alternatives were provided from the adviser's personal experience with similar or identical decisions. This observable fact explains the mechanism of how HOSN is used as a support tool for the DM process, where the majority of decision information in HOSN is based on someone else's experience, a solved decision of others, or already answered questions. These factors emphasize an ability to make a decision through use of information from HOSN. When the adviser provides his own experience in a textual description, he follows all the phases of the DM process in a more organized sequence. Through this observed trend, the limitations of literature on DM such as ‘reification’ [42] have been understood in real-life examples. A majority of the DM models and processes are from the perspective of ‘final decision’ and follow the phases of the DM process in sequence, based on the logic and analysis of previous decisions and the rationality of the decision-maker. But if the decision is viewed from the perspective of ‘initial issue’ such as conversations between seekers and advisers in HOSN, then the phases of the DM process do not have a sequence and do not follow any logical process. The second new step of the DM process that we discovered through observation of HOSN is ‘emotional support’ expressed from advisers to seekers in health-related conversations. This step of the DM process explains why people in health decisions turn to online versus offline

114

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

discussions. The emotional support and empathy provided by users who might have experienced similar problems or concerns is weighted higher than support from friends or family who are not necessarily familiar with the issue or have not experienced it. 5.2. Theoretical contributions Foremost, by exploring, understanding and observing various HOSN, we could recognize and scrutinize the DM process of various stakeholders. Our findings provide a controversial insight into the Simon [62] theory of rational DM process and support the Langley et al. [42] and Mintzberg et al. [49] view of anarchical DM process with unstructured sequence (e.g., [40]) of DM phases. This study provides evidence that HOSN supports the DM process, with the most supported phases being intelligence, design and choice. Implementation and monitoring were found to have partial support, feasible only from the aspect of “experience of others.” An analysis of HOSN conversations helped us visualise how HOSN supports the phases of the DM process; the illustrations (Appendices A–D) show that HOSN conversations are heavily concentrated on three phases of DM process: intelligence, design and choice. Conversation flow is randomly allocated within phases of the DM process, and discussion of a decision can start from any phase. The direction of these conversations cannot be identified in advance since people are different in the way they express themselves in the textual representation of DM in online environments. Our study also captured the new emerging aspect of the DM process in HOSN: “experience of others” or “participant experience”. This aspect can be defined as the DM process of already implemented decision(s) that is expressed in the online networks. Experience of others can be re-used by different stakeholders for the same or similar decision-making situations. This additional step of the DM process was found to be a supportive tool for stakeholders searching for information, advice, models, or alternatives that can support the DM process. One of the contributions of this study is an understanding of the decision-making process from the perspective of HOSN users. This study synthesised the theoretical view on DM, DM process, and HOSN with findings from the exploratory and explanatory research studies into key artefacts in the form of supported hypotheses. The findings provide a new way of thinking and understanding of HOSN in the support of phases of DM process that fulfil decision maker requirements. The stakeholders use HOSN primarily according to their DM purpose. The information gained from HOSN for DM tends to be distorted by human cognitive biases. The HOSN stakeholder in the process of DM is characterized by their DM style and their participation style. DM style in the use of HOSN was found to have little influence on the support of the phases of the DM process, in contrast to participation style, which showed that advisers and seekers use HOSN to support more phases of the DM process than observers. Using netnography as the qualitative research methodology, this study also provided a detailed practical guidance for researchers on its use in other studies oriented towards HOSN. We found netnography to be a good fit for studying sensitive topics like health and providing insights into difficult to investigate subjects of research like the DM of stakeholders. 5.3. Practical implications In addition to evaluating the validity of rational and anarchical models of decision-making in HOSN, this study was also motivated to show how HOSN is capable of supporting the DM process and some of its phases. The conversations that evidenced support by HOSN have been extracted and presented visually in Appendices A–D. Conversations tend to follow an unstructured process of DM. The question that is raised by this finding is how HOSN should be designed in order to support the anarchical process of DM. However, there is still no evidence

that unstructured conversations positively reflect the support function of HOSN; it is likely that there is a need to structure the conversations according to phases of DM process. Structuring the conversations provides an opportunity to efficiently monitor the content of HOSN. Decisions can be practically executed based on the information provided on the walls of the HOSN. Therefore, it can be assumed that a vast number of people make their personal or professional decisions based on the information provided in HOSN. This statement automatically raises the question of quality of information provided online for DM. The freedom of making choices based on information from HOSN can have a negative impact on stakeholders. The quality of provided recommendations by other members of HOSN is unknown. Imperfect information can lead to inefficient allocation of decision-making resources that influence the outcome of the chosen alternatives. Therefore the governance of published information should be considered by HOSN providers, especially if the network aims to be used for efficient decision support. Designers need to be aware of these factors in order to provide efficient support to stakeholders. By providing efficient information search and expert opinion, the quality of the perceived information can be raised, which leads to a better quality of implemented decisions. In terms of information accuracy for healthrelated or any other topic-oriented information, we advise encouragement of the development of specific search engines that are oriented to the topic of search. In the case of HOSN, it should be highlighted that it is also necessary to periodically review the healthcare websites and refer HOSN information seekers only to selective sites that are capable of providing up-to-date information. The HOSN providers should also address the question of how to provide a stable and trustful environment in order to keep existing participants and attract new members. As stated in the analysis and discussions of findings, the availability of expert opinion on any type of networks is valuable. Problems can arise in the area of communication between patients and doctors through HOSN. Regardless, there are many measurable advantages from e-patients' perspectives, but for doctors there is little benefit and even more stress and legal obligations in providing advice on public HOSN. From a practical perspective, the developers of HOSN should ensure that experts' opinions exist in HOSN discussions. Also advised is the provision of some beneficial advantage to the experts that would encourage their participation and attract them to freely share their experience and knowledge. One of the benefits that doctors can gain by participation in HOSN is a reputation or awareness around epatients that can be spread by electronic word of mouth. The aspect of professional users opens up the question of how to ensure that professionals also can find support for their DM process. By observation in HOSN, it was identified that professional networks (doctors to doctors or health researchers to doctors) have found its usefulness through the ability to share information and experience with each other that can affect the well-being of patients. Furthermore, by use of HOSN, doctors are able to better understand their patients outside their appointment visits. From an observation of HOSN and analysis of findings, the new phenomenon of ‘emotional support’ has been distinguished in the DM conversations. This might not find its existence in other networks, but on the subject of health it has been found as an occurrence of kindness and attention of HOSN members to one another. The design and structure of the HOSN website should consider means to encourage expressions of emotional support where appropriate and necessary. The quality and trustworthiness of HOSN and the openness of health professionals to the generation and use of information found at HOSN have been found to positively affect the relationship between patients and health care providers (e.g., [39]). The future is in online technologies and development of healthcare strategies that can ensure that patients and doctors are informed about innovations, and that both can benefit from online communications. Therefore it is important to

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

address the issues of abilities to participate in HOSN by doctors and patients. 5.4. Possible extensions We are thankful to one of the reviewers for pointing out an important data component of online social networks that is missing in this study — that of real-time chatrooms as well as personal messages. Having said that, our choice of forums as data source for our analyses was

115

intentional. Our reasoning was that, unlike chatrooms where instantaneous response was expected of the participants in real-time, forums allowed the participants enough time to think through before posting something online. We believe that the longer temporal component elicits the intended decision making stages/sequence since it drastically reduces the propensity for knee-jerk reactions. A drawback with forums is the possibility of missing decision-making components — specifically that of the final decisions. Given these, an extension to this study would be the inclusion of real-time health-related conversations in chatrooms.

Appendix A. Cancer network The table shows the full process of how netnography was employed in this study. CP-02 – signifies Cancer Post # 2; the title of the online thread is shown to provide more background “How this therapies can treat cancer! herbal medicines for cancer, yoga for cancer, acupuncture for cancer?” – the title indicates the concern or decision that the seeker is looking to find answers for. (NB: the exact title is copied from the conversation, and has not been modified or grammatically altered). The notes section provides background on the post, the URL; the statistical information, question type and main question. Columns indicate who (seeker or adviser) is initiating the post and the DM phases. Then the conversation is coded by the coding system (shown below).

New steps of DM process

EX—exit, E—entry, ES—emotional support adv—advertisement, educt—educational information, bcgr—background AE—adviser entry SE—seeker entry IA—intelligence by adviser IS—intelligence by seeker AM—adviser model SM—seeker model AMA—adviser model based on his experience SC—seeker choice AO—adviser option SI—seeker implementation AA—adviser alternative MS—monitoring seeker CA—choice by adviser AI—adviser implementation MA—monitoring by adviser

New categories of DM phases

Each post is also identified if it can be categorised as educational (information source) or just advertisement (have a commercial perspective that is provided by advisers).

CP-02 How this therapies can treat cancer! herbal medicines for cancer, yoga for cancer, acupuncture for cancer? Notes URL: http://www.drug3k.com/forum_health/Cancer/Howthis-therapies-can-treat-cancerherbal-medicines-for-cancer-yoga-for-canceracupunture-for-cancer-63313.htm Post statistics 2410 words, 10 messages, 10 authors Purpose of post Information and Educational purpose Question type and purpose: Question is seeking for answers, models, decision and options. Seeker is looking for experience of other who had tried these therapies for cancer. Main question: How?

Seeker/adviser Post 1 — Seeker 1 Post 2 — Adviser 1

Intelligence 1.IS 1.IA

Post 3 — Adviser 2 Post 4 — Adviser 3

2.IA 1.IA 3.IA(educt)

Post 5 — Adviser 4

Post 6 — Adviser 5

1.IA

Post 7 — Adviser 6 Post 8 — Adviser 7 Post 9 — Adviser 8

1.IA 5.IA(educt) 1.IA(educt) 4.IA

Post 10 — Adviser 9

1.IA(educt)

Design model 2.AA (adv) 1.AO 2.AO, AM 5.AM 1.AM 3.AA 6.AO, ES 2.AO (adv)

Choice

Implementation

Monitoring

4.CA (adv) 2.CA

4.IA

5.MA

3.IA

4.MA

2.CA 6.CA 2.AO, AM 1.AM 3.AM

2.CA

It can be seen that conversations start at different phases and their sequence can be observed by the sequence of numbers. For example, Post 5 in the table above is read as follows:

Seeker/adviser Post 5 — Adviser 4

Intelligence

Design model

Choice

Implementation

Monitoring

1.AM 3.AA 6.AO, ES

2.CA

4.IA

5.MA

An adviser started the conversation from the design phase by providing his own adviser model (AM), then he moves into the choice phase where he states the choice he would make or made already (CA) (depends on the content of the conversation), then the conversation goes back to the design phase to review existing or new model or alternatives available (AA) which is also provided by advisers; after the adviser shares possible implementation of advised model (IA) — where and how the action should take place (based on previous experience) and then provides insights on the

116

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

consequences of the decision made from the implementation phase by monitoring it (MA). The conversation ends at the design phase where adviser concludes the conversation by providing possible options (AO) as well as some sort of emotional support (ES) that is addressed to a seeker or other participants of the conversation.

Appendix B. Weight loss network WLP-03 Whey protein vs. creatine for weight loss Notes URL: http://tracker.dailyburn.com/forums/supplements/topics/ whey_protein_vs_creatine_for_weight_loss Post statistics 2241, 21 messages, 21 authors Purpose of post Gain Information for an Implementation Question type and purpose: Seeker is looking for advice, solution or further actions Find an answer or model and make a choice Main question: What is better for fat loss, creatine or whey protein ? Did anyone feel….? Is it true? Seeker is looking for answer and advice Seeker is looking for advice/answers

Seeker/adviser Post 1 — Seeker 1 Post 2 — Adviser 1 Post 3 — Adviser 2

Intelligence 2.IS 2.IA 4.IA(educt)

Post 4 — Adviser 3 Post 5 — Adviser 4 Post 6 — Adviser 5

2.IA(educt)

Post 7 — Adviser 6 Post 8 — Adviser 7 Post 9 — Adviser 8 Post 10 — Adviser 9

1.IA 4.IA 1.IA

Post 11 — Adviser10

1.IA 1.IA (educt) 3.IA 5.IA 3.IA

Post 12 — Adviser11

1.IA

Post 13 — Seeker 2 Post 14 — Adviser12

1.IS, EX 4.IA 7.IA 1.IA (bcgr) 8.IA 1.IA 3.IA 5.IA (adv) 1.IA 6.IA, EX 1.IS (bcgr) 4.IS

Post 15 — Adviser13 Post 16 — Adviser 14

Post 17 — Adviser 15 Post 18 — Seeker 3 Post 19 — Adviser 16

Post 20 — Adviser 17 Post 21 — Adviser 28

1.IA (adv) 2.IA (adv)

Design model 1.SM 1.AA 1.AO AMA 1.AO, AM 1.AM 2.AM

Choice

3.AC

2.AA 1.AO, AMA

2.AC

Implementation 3.SI

Monitoring 4.MS

2.AI

3.MA

3.AI

4.MA

3.AC

2.AM 4.AM

6.AC

1.AO, AMA 2.AM AMA 6.AO

2.AC 5.AC

3.AI

4.MA

1.AMA 6.AA 2.AMA

5.AC

2.AI

3.MA

3.AC 9.AC 6.AC

4.AI 6.AI 7.AI

5.MA 7.MA 8.MA

3.AI

4.MA

3.AI

4.MA 6.MA

2.AM 4.AM 2.AM 5.AO 3.SM 1.AM, AO 2.AMA 5.AO 2.AMA 1.AM

2.SC 7.AC

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

117

Appendix C. General health network GHP-01 Does anyone have any experience with having sarcoidosis in the lungs? Notes URL: http://www.imedix.com/questions/q116573-anyone_ experience_having_sarcoidosis_lungs Post statistics 525 words, 7 messages, 7 authors Purpose of post Information purpose Question type and purpose: Seeker is looking for adviser who has the same diagnosis; searching for answers and information on his health concern Main question: Does anyone have any experience..?

Seeker/adviser Post 1 — Seeker 1 Post 2 — Adviser 1 Post 3 — Adviser 2 Post 4 — Adviser 3 Post 5 — Adviser 4 Post 6 — Adviser 5 Post 7 — Adviser 6

Intelligence 1.IS 1.IA(educt) 1.IA

Design model 2.SM 2.AM, AO 1.AM, AMA, ES 2.AM 1.AMA 3.AO 1.AMA 1.AMA

Choice

Implementation

Monitoring

2.AI

4.AC

3.AI

2.MA, ES 2.MA 5.MA

118

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

Appendix D. Plastic surgery network

PS-04 Pics of me: Help me choose procedures as I'm butt ugly! Notes URL: http://messageboards.makemeheal.com/cheek-implants-lift/picshelp-choose-procedures-butt-ugly-t139107.html Post statistics 1014 words, 8 messages, 6 authors Purpose of post Make a choice/decision Question type and purpose: Seeker providing pictures of herself and asking opinion of others on what type of plastic surgery procedures she should undertake. Main question: Which procedures would be best for me?

Seeker/adviser Post 1 — Seeker 1 Post 2 — Adviser 1 Post 3 — Seeker 1 Post 4 — Adviser 2 Post 5 — Adviser 3 Post 6 — Adviser 4 Post 7 — Adviser 5 Post 8 — Adviser 6

Intelligence 1.IS (bcgr) 1.IA 2. IS (bcgr), EX – 4.IA, ES 1.IA 4.IA

Design model 2.SM 3.AM 1.SM – 1.AM, AO 3.AM 3.AM 8.AM 2.AM 1.AO

Choice

Implementation

Monitoring

–

–

5.AI

6.MA

4.AI

3.MA 2.MA 5.MA

2.AC – 2.AC 2.AC 7.AC 1.AC 3.AC

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

References [2] A.L. Barabási, Network medicine — from obesity to the “diseasome”, New England Journal of Medicine 357 (4) (2007) 404–407. [3] L.R. Beach, Broadening the definition of decision making: the role of prechoice screening of options, Psychological Science 4 (4) (1993) 215–220. [4] C.K. Bichakjian, J.L. Schwartz, T.S. Wang, J.M. Hall, T.M. Johnson, J. Sybil Biermann, Melanoma information on the internet: often incomplete — a public health opportunity? Journal of Clinical Oncology 20 (1) (2002) 134–141. [5] D.M. Boyd, N.B. Ellison, Social network sites: definition, history, and scholarship, Journal of Computer-Mediated Communication 13 (1) (2008) 210–230. [6] K. Byström, K. Järvelin, Task complexity affects information seeking and use, Information Processing & Management 31 (2) (1995) 191–213. [7] R. Chen, Living a private life in public social networks: an exploration of member self-disclosure, Decision Support Systems 55 (3) (2013) 661–668. [8] N.A. Christakis, Social networks and collateral health effects, BMJ 329 (7459) (2004) 184–185. [9] R.J.W. Cline, K.M. Haynes, Consumer health information seeking on the Internet: the state of the art, Health Education Research 16 (6) (2001) 671–692. [10] M.D. Cohen, J.G. March, et al., A garbage can model of organizational choice, Administrative Science Quarterly 17 (1) (1972) 1–25. [11] T. Comsti, “Fox Shows How Internet is Changing Health Care” in Peer-To-Peer Health Care Retrieved from http://nihrecord.od.nih.gov/newsletters/2011/09_16_2011/ story2.htm 2011. [12] S. Cooke, N. Slack, Making Management Decisions, Hall Europe, Great Britain, 1991. [15] H. Dreyfus, S. Dreyfus, Mind Over Machine: The Power of Human Intuition and Expertise in the Era of the Computer, Free Press, New York, 1986. [16] W. Duan, B. Gu, A.B. Whinston, Do online reviews matter? — An empirical investigation of panel data, Decision Support Systems 45 (4) (2008) 1007–1016. [17] G. Ellingsen, E. Monteiro, The organising vision of integrated health information systems, Health Informatics Journal 14 (3) (2008) 223–236. [18] G. Eysenbach, J. Powell, M. Englesakis, C. Rizo, A. Stern, Health related virtual communities and electronic support groups: systematic review of the effects of online peer to peer interactions, BMJ 328 (7449) (2004) 1166. [19] T. Ferguson, G. Frydman, The first generation of e-patients, BMJ 328 (7449) (2004) 1148–1149. [20] V. Fox, Internet Study Finds Most Americans Get Their Answers from The Internet, http://searchengineland.com/pewInternet-study-finds-most-americans-get-theiranswers-from-the-Internet-13028 2007. [21] S. Fox, “Medicine 2.0” in Peer-to-peer Healthcare, http://pewInternet.org/Reports/ 2011/Medicine-20/Part-1.aspx 2011. [22] N. Fox, C. Roberts, GPs in cyberspace: the sociology of a ‘virtual community’, The Sociological Review 47 (4) (1999) 643–671. [23] S. Fox, J. Brenner, “Family Caregivers Online”, Pew Internet: A Project of the Pew Research Center, http://pewInternet.org/~/media//Files/Reports/2012/PIP_Family_ Caregivers_Online.pdf 2012.

119

[26] B.G. Glaser, A. Strauss, The Discovery of Grounded Theory: Strategies for Qualitative Research, Aldine, Chicago, 1967. [27] D. Godes, D. Mayzlin, Using online conversations to study word-of-mouth communication, Marketing Science 23 (4) (2004) 545–560. [30] G.J. Hart, The five W's: an old tool for the new task of audience analysis, Technical Communication 42 (2) (1996) 139–145. [31] D. Hobbs, Ethnography, in: V. Jupp (Ed.), The Sage Dictionary of Social Research Methods, SAGE, London, 2006. [32] HSAfinder.com, Taboo Health Topics Are Popular on the Web, http://www. hsafinder.com/Taboo-Health-Topics-Are-Popular-on-the-Web 2008. [34] H.S. Kim, S.S. Sundar, Using interface cues in online health community boards to change impressions and encourage user contribution, Paper Presented at the Proceedings of the 2011 Annual Conference on Human Factors in Computing Systems, Vancouver, BC, Canada, 2011. [35] H.K. Klein, M.D. Myers, A set of principles for conducting and evaluating interpretive field studies in information systems, MIS Quarterly 23 (1) (1999) 67–93. [36] R.V. Kozinets, The field behind the screen: using netnography for marketing research in online communities, Journal of Marketing Research 39 (1) (2002) 61–72. [37] R.V. Kozinets, Netnography: Doing Ethnographic Research Online, Sage Publications, 2010. [38] R. Kumar, J. Novak, A. Tomkins, Structure and Evolution of Online Social Networks Link Mining: Models, Algorithms, and Applications, in: P.S.S. Yu, J. Han, C. Faloutsos (Eds.), Springer, New York, 2010. [39] B.M. Lagan, M. Sinclair, W.G. Kernohan, Internet use in pregnancy informs women's decision making: a web-based survey, Birth 37 (2) (2010) 106–115 (June). [40] S.D. Lambert, C.G. Loiselle, Health information-seeking behavior, Qualitative Health Research 17 (8) (2007) 1006–1019 (October). [41] R. Langer, S.C. Beckman, Sensitive research topics: netnography revisited, Qualitative Market Research: An International Journal 8 (2005) 189–203. [42] A. Langley, H. Mintzberg, P. Pitcher, E. Posada, J. Saint-Macary, Opening up decisionmaking: the view from the black stool, Organisation Science 6 (3) (1995) 260–279. [43] V. Lendel, The decision-making about the selection of CRM information system, Journal of Information Technology Control and Management Systems 6 (1) (2008) 85–93. [44] W. Macias, L.S. Lewis, V. Shankar, Dr. Mom and Dr. Web: a qualitative analysis of women's use of health information on the web, Journal of Interactive Advertising 4 (2) (2004) 9–19. [46] G.E. Marcus, Ethnography in/of the world system: the emergence of multi-sited ethnography, Annual Review of Anthropology 24 (1995) 95–117. [47] S. Mathur, N. Shanti, M. Brkaric, V. Sood, J. Kubeck, C. Paulino, Surfing for scoliosis: the quality of information available on the internet, Spine 30 (23) (2005) 2695–2700. [48] M.B. Miles, M.A. Huberman, Qualitative Data Analysis: An Expanded Sourcebook, 2nd ed.Sage Publications, Thousand Oaks, CA, 1994. [49] H. Mintzberg, D. Raisinghani, A. Théorêt, The structure of “unstructured” decision processes, Administrative Science Quarterly 21 (2) (1976) 246–275.

120

V. Sadovykh et al. / Decision Support Systems 74 (2015) 102–120

[50] S. Mnookin, The Huffington Post and the ongoing fear that vaccines might cause autism, Scientific American (2011) http://blogs.scientificamerican.com/guest-blog/ 2011/02/11/the-huffington-post-and-the-ongoing-fear-that-vaccines-might-causeautism/. [52] M. Myers, Investigating information systems with ethnographic research, Communication of the AIS 2 (4) (1999) 1–20. [53] S.A. Orlovsky, Decision-making with a fuzzy preference relation, Fuzzy Sets and Systems 1 (3) (1978) 155–167. [54] Pew Internet Project, Online Health Research: 8 in 10 Web Users Look for Online Health Data available at http://www.marketingcharts.com/direct/8-in-10-webusers-look-for-online-health-data-16107/pew-research-online-health-researchby-age-feb11gif/ 2011. [56] L. Rainie, Online Health Seeking: How Social Networks Can be Health Communities available at http://www.slideshare.net/PewInternet/online-health-seeking 2010. [57] D. Rodan, L. Uridge, L. Green, Using nicknames, pseudonyms and avatars on HeartNET: a snapshot of an online health support community, Paper Presented at the ANZCA Conference, Canberra, Australia, 2010. [58] V. Sadovykh, D. Sundaram, S. Piramuthu, Do online social networks support decision-making? Decision Support Systems 70 (2015) 15–30. [59] A.P. Sage, Information Processing in Individuals and Organizations Decision Support Systems Engineering, John Wiley & Sons Inc. (US), 1991. [60] M. Schifrin, H. Wolinsky, Use With Care, http://www.forbes.com/best/2001/0625/ 026.html 2001. [62] H.A. Simon, Theories of decision-making in economics and behavioural science, The American Economic Review 49 (3) (1959) 253–283. [63] H.A. Simon, Decision-making and problem solving, Report of the Research Briefing Panel on Decision-making and Problem Solving, National Academy of Sciences, Washington, DC, 1986. [64] M. Sinclair, N.M. Ashkanasy, Intuition, Management Learning 36 (3) (2005) 353–370. [65] S. Spiggle, Analysis and interpretation of qualitative data in consumer research, Journal of Consumer Research 21 (3) (1994) 491–503. [66] Sutterhealth.org, Popular Health Topic, sutterhealth.org/yourhealth/health_topics/ health_topics.html 2010. [67] M. Swan, Emerging patient-driven health care models: an examination of health social networks, consumer personalized medicine and quantified self-tracking, International Journal of Environmental Research and Public Health 6 (2) (2009) 492–525. [68] Y.-J. Tu, W. Zhou, S. Piramuthu, Identifying RFID-embedded objects in pervasive healthcare applications, Decision Support Systems 46 (2) (2009) 586–593.

[69] T. Verbraken, F. Goethals, W. Verbeke, B. Baesens, Predicting online channel acceptance using social network data, Decision Support Systems 63 (2014) 104–114. [70] S. Vodanovich, D. Sundaram, M. Myers, Digital natives and ubiquitous information systems, Information Systems Research 21 (4) (2010) 711–723. [71] K.J. Ward, Cyber-ethnography and the emergence of the virtually new community, Journal of Information Technology 14 (1) (1999) 95–105. [72] M. White, S.M. Dorman, Receiving social support online: implications for health education, Health Education Research 16 (6) (2001) 693–707. [74] W. Zhou, W. Duan, S. Piramuthu, A social network matrix for implicit and explicit social network plates, Decision Support Systems 68 (2014) 89–97. [75] G.P. Huber, R.R. McDaniel, The Decision-Making Paradigm of Organizational Design, 32 (5) (1986) 572–589. [76] K. Malterud, Qualitative research: standards, challenges, and guidelines, The Lancet 358 (9280) (2001) 483–488. [77] G. Eysenbach Gunther, C. Köhler, How do consumers search for and appraise health information on the world wide web, Qualitative study using focus groups, usability tests, and in-depth interviews BMJ 324 (2002) 573–577. Valeria Sadovykh is a PhD student in the Department of Information Systems and Operations Management, Business School, The University of Auckland. Her research interests include use of Online Social Networks as a support tool for Decision Making Process, Online Decision Support Systems, Netnography and Content Analysis of Online Social Media. David Sundaram has a varied academic (Electronics & Communications, Industrial Engineering, and Information Systems) as well as work (systems analysis and design, consulting, teaching, and research) background. His primary research interests are in the design and implementation of adaptive systems that are flexible and evolvable and support informational, decisional, knowledge and collaboration needs of organizations and individuals. Related research interests that support the above include (1) adaptive organizations and supporting them through the interweaving of deliberate and emergent strategies (2) process, information, decision, and visualisation modeling (3) sustainability modeling and reporting (4) ubiquitous information systems (5) enterprise systems and their implementation. Selwyn Piramuthu is a Professor in the Information Systems and Operations Management Department at the University of Florida. His research interests include online social networks, RFID systems, pattern recognition and its application in supply chain management, computer-aided manufacturing, and financial credit-risk analysis.