=Paper=
{{Paper
|id=Vol-1276/MedIR-SIGIR2014-01
|storemode=property
|title=Designing for Health Exploratory Seeking Behaviour
|pdfUrl=https://ceur-ws.org/Vol-1276/MedIR-SIGIR2014-01.pdf
|volume=Vol-1276
|dblpUrl=https://dblp.org/rec/conf/sigir/PangVCP14
}}
==Designing for Health Exploratory Seeking Behaviour==
https://ceur-ws.org/Vol-1276/MedIR-SIGIR2014-01.pdf
Designing for Health Exploratory Seeking Behaviour
Patrick Cheong-Iao Pang1,2, Karin Verspoor1,3, Shanton Chang1, Jon Pearce1
1
Department of Computing and Information Systems, The University of Melbourne
2
NICTA Victoria
3
Health and Biomedical Informatics Centre, The University of Melbourne
Parkville, Victoria 3010, Australia
cipang, karin.verspoor, shanton.chang, j.pearce@unimelb.edu.au
ABSTRACT information seeking processes tends become more exploratory
The Internet has become a popular source of health information. [7][8]. We argue that this exploratory search behaviour applies to
However, in-depth understanding of the information seeking health information seekers as well. In contrast to executing a
behaviour of online health information is limited. We conducted search query and reading through the result, exploratory search
an experiment to investigate the information needs and behaviours involves a series of cognitive learning and query reformulation
of health information seekers. This paper reports on a model of processes. A more complete picture of the knowledge domain is
behavioural patterns drawn on the experimental results, and being built in this process. The exploratory process also implies
implications for designing a better user experience for the the existence of both learning and investigative activities. Seekers
exploration of online health information. end up knowing more information than they expected at the
beginning. One of our goals is to design a system that better
supports such exploration, which traditional search engines are
Categories and Subject Descriptors not designed for.
H.1.2 [Information Systems]: User/Machine Systems – Human
information processing. In order to understand more deeply the effects on human
behaviour of these health information needs, we ran interview
sessions and observation experiments with 20 participants. During
General Terms the observation, the participants displayed diverse behavioural
Design, Experimentation, Human Factors. patterns. Summarised from these patterns, we propose a model to
describe the seeking behaviour in terms of research tactics and
Keywords reading engagement. The findings indicate there exists a lack of
Exploratory search, health, information seeking, human behaviour features in both search engines and health websites to support
health information seeking, which is essentially an exploratory
search. We plan to build an experimental health website to
1. INTRODUCTION address these problems in the next phase of our research.
The Internet provides a variety of choices for consumer health
information, from official health websites, private health service
providers, personal blogs, to discussion forums. Studies have 2. LITERATURE REVIEW
shown that lay-people look for health information online when In this section we draw on the literature to explain what we mean
they encounter health problems [1][2]. While we understand the by the three phrases: online health information seeking, health
demands of accessing health information on the Internet, there are information needs and exploratory search. These concepts are
few studies investigating the information needs and the guiding the direction of this research.
characterising information seeking behaviours of these seekers.
Our research aims to bridge this gap. 2.1 Online Health Information Seeking
Health information seeking on the Internet is different from other
Search engines are the de-facto primary method of finding types of searches in many ways. Lay-people usually have only
information on the web. This also applies to health seekers. limited knowledge in the medical domain [9], or face difficulties
However, the user experience in the health context is often not in utilising technical or medical language for searching [4][5]. On
satisfactory due to a number of factors, including the nature of the other hand, dealing with health issues is stressful and
health information, the level of knowledge possessed by the uncertain and very likely to demonstrate a different information
seeker, and the skill of formulating search keywords [3][4][5][6]. need [10].
When dealing with unfamiliar and unknown problems, or A study showed 72% of U.S. Internet users have tried to access
involving a task for which the goal remains unclear, the health information online [2]. Seekers look for a broad range of
health information, which includes disease information, causes
and treatments, diet information, health lifestyles, etc.
[1][2][9][12][13][14][15], and in various stages of a health
problem [16][17]. The diverse types of health information
demanded reflect differences in information needs which result in
Copyright is held by the author/owner(s).
different seeking behaviour.
MedIR 2014, July 11, 2014, Gold Coast, Australia.
5
2.2 Health Information Needs append to a Wikipedia page on urination problems and their
Information needs arise when people realise their existing symptoms. The first scenario was designed with the aim of
knowledge is inadequate to satisfy their goal. Finding information observing seeking behaviours for explicit recognised needs,
is an attempt to bridge a knowledge gap. A knowledge gap whereas the second targeted for unrecognised needs with a more
appears whenever people perceive there is not enough information vague description of a health problem.
in their minds and as a result they will start searching for Interviews were transcribed and reduced to a number of codes
information to fill the blanks [11]. This process is also known as a iteratively [22]. Content that was relevant to health information
sense-making process [18]. seeking behaviours were organised into themes. Themes were
derived with a thematic analysis approach [23]. Screen recordings
Alzougool et al. investigated different information needs in the
were reviewed manually. A navigation graph was built for each
health context [10][19]. They propose that health information
participant to describe the web activity patterns in each session.
needs can be further classified into recognised and unrecognised
needs. Cartright et al. [20] argued that health information seeking
can be split into two partitions – evidence-based and hypothesis- 4. RESULTS
directed. Through analysing queries in search engine sessions, In total 20 participants completed our lab experiment (11 male; 9
they clustered the foci of searches in terms of causes, symptoms, female). In terms of identity, they comprised 8 students, 9
remedies, or combinations of any of these. This could be useful to university staff and 3 external participants. The age distribution is
predict the information needs of exploratory health seekers. listed in Table 1.
Table 1. Age Distribution of Participants
2.3 Exploratory Search
Exploratory search involves learning and investigation in addition Age Group N Percentage
to lookup efforts, where the seeker interacts with information 21-30 8 40%
systems to retrieve a wider range of information [7]. Exploratory 31-40 6 30%
search can be found when the individual tries to address
unfamiliar or unknown problems [21], as may be the case for 41-50 4 20%
health-related concerns. White and Roth add that people who are Over 50 2 10%
unfamiliar with the domain of their goals, or unsure about the
ways to achieve their goals, or even unsure about their goals, will
engage in exploratory search [8].
4.1 Motivation for Exploratory Search
3. RESEARCH METHOD One of the themes emerging from the interviews regarded the
Previous research about health information seeking behaviours motivation for exploratory search. Not every search about health
does not capture well the actual patterns of the interactions among topics is an instance of exploratory search, however, we have
search engines and health websites. The aim of this experiment observed that both recognised and unrecognised information
was to form an in-depth understanding of information needs and needs motivated people to perform exploratory search in our study.
the patterns of related information seeking behaviours. We Firstly, searches triggered by recognised needs are found to be
arranged sessions with individual participants in which we more exploratory. For example, participants with health problems
interviewed them about their experiences on health information or those diagnosed with a certain illness. The need is also
seeking, and then gave them two tasks to carry out. We recorded demanding if the issue is related to people’s loved ones. They had
the interviews and screen activities for further analysis. a clearer mind about what information is helpful for the scenario,
The study was carried out from October to December of 2013. and possessed explicit information needs on aspects such as
Posters were presented in various locations across the university treatments and remedies. In this case, the recognised need mainly
for recruiting participants. E-mail invitations were sent out to has the purpose of helping to understand the complete picture of
encourage participation. Participants were also allowed to invite the situation or getting more options for facing the health problem.
potential participants to our study through their connections. The seeker is persistent in trying different ways to discover and
Recruitment continued until data saturation was achieved. This read information as well.
experiment was voluntary and no incentive was given to
Unrecognised needs were observed to stimulate exploratory
participants for the study.
search as well. People do not have a clear target, and therefore
Each study lasted about an hour and consisted of three sections: tend to approach different sources to make sense of the
the first section was a semi-structured interview about their past information – this is illustrative of exploratory behaviour during
experiences of finding health information on the web; participants the search. Examples include people passively encountering
were then given a computer to find online health information for contradictory messages, suspecting the validity of the information,
two pre-defined tasks in the second part; the last section included or simply feeling curious about certain information. They seek
another semi-structured interview to help researchers further additional sources without knowing what exactly is needed nor
understand how participants performed the tasks. We did not why it is needed. They usually become more open minded to the
restrict the participants to use a specific website or instruct them information obtained but are still cautious about it to avoid wrong
to use a search engine. To avoid bias, we cleared the home page information.
and all browsing history in the browser prior to each session. All
interview content and screen activities were recorded. 4.2 Behavioural Patterns
The search tasks represented two different styles of common Participants were requested to perform two search tasks in our lab
health scenarios. In one scenario, participants were asked to find study sessions. For each task, we manually constructed a
information on how to care for a diabetic family member; in the navigation graph to describe the pattern of the interaction among
other, participants were tasked with identifying information to search engines, individual web pages and clicks on hyperlinks. 36
6
graphs were generated while 4 tasks not completed by participants Seekers in Figure 1(d) use hyperlinks to discover new information.
due to time constraints on these lab sessions. From these graphs They choose a small number of good websites (usually large and
we identified four common patterns (Figure 1). During the search reputable) for examination. In these well-designed health websites,
process of each task, the overall patterns of the individual’s connections among pages are well-defined and hyperlinks are
activities may contain one or more these common patterns shown placed in a useful manner. Seekers can trace the related
in the figure. information through levels of hyperlinks easily and do not need to
query the search engine as often as in other scenarios.
4.3 Model of Behavioural Outcome
From the perspective of enhancing user experience and suggesting
design implications, conceptualising the information seeking
behaviour is crucial. Drawn on the patterns in the previous section,
we build up a model to abstract the behaviour outcome of online
health information seeking behaviour (Figure 2).
Extensive Research Extensive Research
Extensive
Research Tactics
Low Reading High Reading
Basic Research Basic Research
Basic Low Reading High Reading
Low High
Reading Engagement
Figure 2. Model of Behavioural Outcome
This matrix presents a combination of seeker’s research tactics
and reading engagement. Research tactics represent the eagerness
and motivation for finding out in-depth information. For more
exploratory seekers, a wider range of information is needed, and
thus the extensive tactics represent a greater effort to locate, filter,
learn and discover other information within the current knowledge
domain. Whereas less research effort will be put for the basic
tactics, which often appear when looking for surfaced, easy-to-
obtain and easy-to-read information.
Reading engagement measures the duration of reading and the
intention of absorbing the information. Skimming and reading just
the page summary fall into the group of low engagement while
pursuing and digesting the information is considered as high
engagement.
4.4 Guidelines for Design
Figure 1. Typical Patterns of Health Exploratory Search Drawn on the abovementioned findings, we propose two areas of
Figure 1(a) shows the simple pattern of executing a keyword improvements for enhancing the user experience of exploratory
query on a search engine, skimming one of the search results, and health information seekers. These are (1) assisting the discovery
then finishing the information seeking process. This is common of new information, and (2) adapting to users’ reading needs.
when the seeker just wishes to have some quick facts. The search
The discovery of information within a website is important. In
stops once the information is found.
general, health websites collect many articles but we have noticed
Figure 1(b) is similar to (a) in the small number of searches that seekers are not always able to reach all of them. This implies
performed, but the seeker engages in longer and deeper readings. a problem of either users not knowing what they want, or that they
The seeker examines the web page returned by the search engine, cannot effectively use a search engine to explore. As seen in
also follows hyperlinks provided page-by-page and continues Figure 2, seekers with the basic research tactics stay within a
reading. The duration of reading is longer as the seeker digests single website rather than utilising a search engine for to look
and absorbs the information. The series of reading continues until other sites. In this regard, a system that understands their
no more valuable information is found. information needs and recommends relevant information for
further reading is preferable.
Figure 1(c) presents not much reading but more query
reformulations. The seeker in this case relies on the search engine Figure 2 identifies a spectrum of reading engagement, suggesting
to explore new information. Most of the time, he/she tries to skim that a health website needs to adjust to both low and high reading
through the search results and picks up only web pages that are levels. In low engagement behaviour, users prefer to skim and
relevant to the information need. Meanwhile the seeker will adjust quickly read through the articles to determine the usefulness
the query keywords with the information read and submit a new before committing to a longer reading. In this case, an abstract or
query, if the overall search result is not satisfactory. The query summary could be provided for their convenience. On the other
reformulation particularly occurs when the user cannot discover hand, users with high reading engagement may prefer a design
new useful information with visible means (e.g. links) and feels emphasising readability, such as font size, line spacing, section
the information in the website is exhausted. navigation, etc.
7
Extensive research seekers (the upper row in Figure 2) generally [9] Zhang, Y., Fu, W.-T. 2011. Designing consumer health
do not have major problems in finding the information. information systems: what do user-generated questions tell
Corresponding to Figure 1 (c) and (d), they illustrate patterns of us? In Proceedings of the FAC 2011, HCII 2011, LNAI 6780.
putting efforts to discover, locate and filter needed information. [10] Alzougool, B., Chang, S. and Gray, K. 2013. The nature and
They also spend most of the time to judge the relevance of the constitution of informal carers’ information needs: what you
materials, and seek alternatives if the information is not relevant. don't know you need is as important as what you want to
Both of the two design implications would be beneficial to this know. information Research. 18, 1 (Mar. 2013).
type of seekers.
[11] Case, D. O. 2002. Looking for Information: A Survey of
Research on Information Seeking, Needs, and Behaviour.
5. CONCLUSION Academic Press, Lexington.
This paper reports on a lab experiment with the aim to understand
online health information seeking behaviour. From this study, we [12] Andreassen, H. K., Bujnowska-Fedak, M. M., Chronaki, C.
have identified that a variety of information needs that drive E., Dumitru, R. C., Pudule, I., Santana, S., Voss, H. and
exploratory search, and a diversity in behavioural patterns. We Wynn, R. 2007. European citizens' use of E-health services: a
have proposed a simple model using two factors, research tactics study of seven countries. BMC public health. 7, 53 (Apr.
and reading engagement, that is useful for reflecting on user 2007), 1-7.
behaviour and starting to think about how we can design systems [13] Bessell, T. L., Silagy, C. A., Anderson, J. N., Hiller, J. E. and
to provide better support for exploratory behaviours. These Sansom, L. N. 2002. Prevalence of South Australia’s online
findings sight the directions we could work on to improve the user health seekers. Australian and New Zealand J. of Public
experience of health information seekers. Health. 26 (Mar. 2002), 170-173.
We will focus on designing a better environment for exploratory [14] Johnson, J. D. and Meischke, H. 1991. Women's preferences
search in the health context. The next phase of this research is to for cancer information from specific communication
build a testing health website with the goal on assisting the channels. The American Behavioral Scientist. 34, 6 (Jul.
discovery of new information and enhancing reading engagement. 1991), 742.
A larger scale of user study will be launched to gather feedback [15] Nicholas, D., Huntington, P., Gunter, B., Withey, R. and
and evaluate the new design elements in this new website. Russell, C. 2003. The British and their use of the web for
health information and advice: a survey. Aslib Proceedings.
6. ACKNOWLEDGEMENT 55, 5, 261-276.
Patrick Pang is supported by the Australian Federal and Victoria
[16] Rutten, L. J. F., Arora, N. K., Bakos, A. D., Aziz, N. and
State Governments and the Australian Research Council through
Rowland, J. 2005. Information needs and sources of
the ICT Centre of Excellence program, National ICT Australia
information among cancer patients: a systematic review of
(NICTA).
research (1980–2003). Patient Education and Counseling,
57(2005), 250–261.
7. REFERENCES
[1] Fox, S. and Jones, S. 2009. The Social Life of Health [17] Ofran, Y., Paltiel, O., Pelleg, D., Rowe, J. M. and Yom-Tov,
Information. Technical Report. Pew Internet & American E. 2012. Patterns of information-seeking for cancer on the
Life Project. Internet: An analysis of real world data. PLoS ONE, 7(9),
e45921.
[2] Fox, S. and Duggan, M. 2013. Health Online 2013.
Technical Report. Pew Internet & American Life Project. [18] Wilson, T. D. 1999. Models in information behaviour
research. Journal of Documentation, 55, 3, 249-270.
[3] Zhang, Y. 2011. Exploring a web space for consumer health
information: implications for design. In Proceedings of [19] Alzougool, B., Chang, S. and Gray, K. 2008. Towards a
iConference 2011 (Seattle, WA). comprehensive understanding of health information needs.
electronic Journal of Health Informatics, 3, 2.
[4] Keselman, A., Browne, A. C. and Kaufman, D. R. 2008.
Consumer health information seeking as hypothesis testing. [20] Cartright M.-A., White, R. W., and Horvitz, E. 2011.
JAMIA. 15, 4 (Jul. 2008), 484-495. Intentions and attention in exploratory health search. In
Proceedings of the 34th International ACM SIGIR
[5] Chapman, K., Abraham, C., Jenkins, V. and Fallowfield, L. Conference on Research and Development in Information
2003. Lay understanding of terms used in cancer Retrieval. SIGIR’11. 65-74.
consultations. Psycho-oncology. 12, 6 (Sep. 2003), 557-566.
[21] Pearce, J., Chang, S., Kennedy, G., Ely, R. B. W. and Ainley,
[6] Luo, G., Tang, C., Yang, H. and Wei, X. 2008. MedSearch: M. 2012. Search and explore: more than one way to find
A specialized search engine for medical information what you want. In Proceedings of the 2012 Australian
retrieval. In Proceedings of 17th ACM Conference on Computer-Human Interaction Conference (Melbourne,
Information and Knowledge Management (Napa Valley, CA, Victoria, Australia, November 26-30, 2012). OzCHI '12.
October 26-30, 2008). CIKM '08.
[22] Creswell, J. W. 2002. Educational Research: Planning,
[7] Marchionini, G. 2006. Exploratory search: from finding to Conducting, and Evaluating Quantitative and Qualitative
understanding. Communications of the ACM. 49, 4 (Jun. Research. Merrill Prentice Hall, Upper Saddle River, NJ.
2006), 41-46.
[23] Braun, V. and Clarke, V. 2006. Using thematic analysis in
[8] White, R. and Roth, R. 2009. Exploratory Search: Beyond psychology. Qualitative Research in Psychology, 3, 2 (Jan.
the Query-Response Paradigm. Morgan and Claypool, San 2006), 77-101.
Rafael, CA.
8