Towards Supporting Awareness for Content Curation: The case of Food Literacy and Behavioural Change Roberto Martinez-Maldonado, Theresa Anderson, Simon Buckingham Shum and Simon Knight Connected Intelligence Centre, University of Technology Sydney Chippendale, NSW, 2007 {Roberto.Martinez-Maldonado,Theresa.Anderson, Simon.BuckinghamShum,Simon.Knight}@uts.edu.au ABSTRACT such information (e.g. showing if most webpages that are being This paper presents a theoretical grounding and a conceptual curated correspond to blog posts rather than evidence-based proposal aimed at providing support in the initial stages of articles, or whether they include references to scientific papers, sustained behavioural change. We explore the role that learning media reports, journals, etc). analytics and/or open learner models can have in supporting life- Another recent movement aimed at raising awareness about long learners to enhance their food literacy through a more personal wellbeing is quantifying different aspects of a person’s informed curation process of relevant-content. This approach daily activity using self-trackers [8]. However, these technologies grounds on a behavioural change perspective that identifies i) have shown considerable limitations in sustained usage [21] (e.g. knowledge, ii) attitudes, and iii) self-efficacy as key factors that users have shown low levels of long-term engagement using smart will directly and indirectly affect future decisions and agency of wearables). As a result, learning about individuals’ best practices life-long learners concerning their own health. The paper offers to promote and maintain their own health can be very challenging some possible avenues to start organising efforts towards the use without community support or effective technological guidance, of learning analytics to enhance awareness in terms of: knowledge or both. Moreover, it makes it harder for an individual to gain the curation, knowledge sharing and knowledge certainty. The paper knowledge necessary to make initial progress towards a sustained aims at triggering discussion about the type of data and behaviour change. We illustrate the value of our approach towards presentation mechanisms that may help life-long learners set a supporting awareness in this context. stronger basis for behavioural change in the subsequent stages. In this paper, we explore the positive role that learning analytics Keywords and/or open learner models can have in supporting lifelong Information curation; Food literacy; Behavioural change; learners in performing a more informed curation process of Learning analytics, OLM’s relevant content. We aim to achieve this by enhancing learners’ awareness in three areas: the knowledge curation process; aspects 1. INTRODUCTION of their collaborative learning process and knowledge sharing; and The proliferation of mobile devices and internet access has the types of sources, which can enhance knowledge trust or provided the means for individuals and communities to have mistrust. The paper presents a theoretical grounding and a access to a wide range of information. Searching and curating conceptual proposal aimed at providing support in the initial information from the Internet have been increasingly identified as stages of sustained behavioural change for the particular case of a popular source for people seeking information about how to take food literacy support. In this paper, we use the term food literacy care of their own health [7; 20]. Users commonly use search to refer to the individual or collective understanding about food engines and visit multiple websites to find information [11]. and nutrition that can empower people to manage their own health However, the types and quality of these sources can vary widely, choices [28]. Our approach is grounded in a behavioural change and can often be contradictory or overwhelming. When a regular perspective that identifies i) knowledge, ii) attitudes, and iii) self- user finds interesting articles, he or she can save it for later use. efficacy as key factors that will directly and indirectly affect However, many times it is hard to keep track of interesting future decisions and agency of life-long learners concerning their content as time passes and, more important, to interrelate and own health. make sense of a number of content sources around a common topic. Information overload [15] and lack of credibility indicators The rest of the paper presents first, the theoretical grounding for [7] have been reported as important factors that can lead to a the application of learning analytics and learner modelling for higher degree of uncertainty and misguidance. Moreover, the collaborative content curation. Next, we define the context of food majority of people do not consistently check the source and date literacy and behavioural change, before presenting our proposed of the health information found online [11]. A number of web conceptual approach along with some initial learning analytics content curation tools (also known as social bookmarking tools) ideas. We conclude with a discussion of future avenues of this are popular solutions for this problem since they commonly allow project. collaboratively annotating, archiving and bookmarking webpages [23]. These may facilitate the organisation of content, information 2. THEORETICAL GROUNDING and knowledge for lifelong learning or for researching about 2.1 Learning Analytics or OLM’s particular topics [9; 18]. However, these solutions do not Learning analytics is a novel and quite holistic perspective that necessarily make evident to people how knowledge is aims to provide support to the various stakeholders of the individually/collaborative built or the meta-information about the educational practice by exploiting data related to learning, sources of the information that could enhance their certainty on teaching or the management of the educational process [25]. A Copyright © 2016 for the individual papers by the papers' authors. Copying permitted only for private and academic purposes. This volume is published and copyrighted by its editors. LAL 2016 workshop at LAK '16, April 26, 2016, Edinburgh, Scotland. distinctive aspect of learning analytics is its emphasis on collaborative content curation, namely, the initial stages that may connecting the collection, analysis and reporting of data about lead to sustained behavioural change in food literacy. learning and its contexts with high quality practical pedagogical approaches [19]. Thus, learning analytics mainly focuses on 3.1 Food Literacy leveraging human judgement by empowering learners and Food literacy (or nutrition literacy) is an emerging term itself that educators with key information about the learning process that can can be described in words of Vidgen and Gallegos [28] as what help them take better and informed decisions [26]. In technical individuals and communities know and understand about food and terms, data is commonly delivered to students, teachers, etc. how to use it to meet their particular needs. In other words, it through visualisations, graphs, notifications, etc. considers the challenge of making healthy food choices as an educational problem. The food literacy model by Fordyce- Research into feeding back traces to learners of their own data is Voorham [10] identifies relationships between three main not new. There has been substantial research and development on elements: individual (which refers to the personal decision Open Learner Models (OLM) [4]. While a learner model making, management of actions and learning about oneself), corresponds to a structured data model constructed from the traces cultural (which takes into account cultural preferences and food of interaction between a learner and a learning system or systems, security) and macro-system (which accounts for the impact of the Open LM’s are designed to be viewed or accessed in some way by environment on food decisions and ethical choices) dimensions. the learner, or by other users (e.g. educators, peers, etc.). Even though there are key differences between learning analytics and Mobile learning practices and technologies offer promising OLM perspectives [14] (e.g. the data that is fed back to students is avenues for articulating a pedagogy for food literacy issues commonly less processed from a LA perspective, or the key role through a social learning approach. Collaborative content curation of adaptation from an OLM perspective), the common aim of both platforms may allow individuals to seek and share information approaches remains the same: to make learners data visible to help with their community (as we aim to achieve). Other mobile them gain understanding of different aspects of their learning [14]. solutions have been used to support individuals to generate content socially situated and connected with the local food We aim to take a stance on both perspectives, in particular, OLMs growers and farmers (e.g. community recipes [12]). In these cases, could be considered as a specific type of learning analytics [14]. a learner model could be generated to gather key information We aim to make visible to learners data about their collaborative about the individual’s actions, the content consulted, and learning information curation process. This may require tuning the type of gained by interacting with different mobile apps (the individual support according to the learners’ particular needs, interests or element of Fordyce-Voorham’s Food Literacy Model). The knowledge (a learner model), delivering visualisations or pushing challenge would be how to account for the social, cultural and notifications and/or recommendations to the learners. systemic factors that can affect individual’s learning and their decision making process (the cultural and macro-system elements 2.2 Collaborative Content Curation of Fordyce-Voorham’s Food Literacy Model). Content curation refers to the activities related to searching, selecting, organising, validating, maintaining, and preserving 3.2 Behavioural Change existing content artefacts [23]. Content curation communities have Besides supporting content curation, we aim to situate our project emerged in parallel to the growth of the generation of web to support learners, at least, on their initial steps towards a content. Different automatic or semiautomatic tools have been sustained behavioural change based on their improvements on developed to support content curation in a range of areas, from food literacy. Behavioral change is a central objective in public scientific content curation communities engaged in solving health, with the main aim of preventing disease [5]. The main complex problems that require many resources, to individuals reason to stand on behavioural change is that it may allow long- curating information for personal use [23]. Some cloud based term adherence to healthy lifestyles (e.g. improving eating habits tools allow creating collections of web resources to keep and physical activity), rather than just the achievement of tasks or individuals’ knowledge organised or to be shared with the goals (e.g. just weight loss). community. Two particular examples that we consider in this paper are the commercial tools Diigo1 and Declara2. There has There are a vast number of theories of behavioural change [1]. been some interest in using these tools as meta-cognitive tools [9; Particularly for health promotion, it has been identified that the 18]. Analysing the logs of learners’ activity collaboratively community and social dynamics play a crucial role in addressing curating data may show traces of the learning process and the way the resistance to change [5]. As a result, our educational strategies the knowledge that is obtained from the curated content develops. for behavioural change in food literacy should be designed within Additionally, social analytics could be applied to the logged data the cultural context using people’s own beliefs (rather than since learners commonly use shared tag names to mark and share imposing an agenda or content on them), with a combination of resources with other people, comment on others’ findings or even well-grounded sources of knowledge and local practices. discuss on particular extracts of the curated web-pages. Our approach is grounded on Meinhold et al.’s approach [22] of Consequently, these tools can become meaningful learning behavioural change by supporting individuals’: resources that provide a social dimension for learning or for general collaborative content curation.  Knowledge: factual claims about the context, which can be individually and collectively shared in virtual communities. 3. CONTEXT This includes the personal knowledge, the environmental In this section, we briefly describe an application context for knowledge (what other people ‘know’) and also formal learning analytics and/or open learner models to support sources of knowledge (articles, publications, research papers). In terms of content curation, the degree of knowledge certainty would be crucial for scaffolding 1 https://www.diigo.com/ behavioural change. 2 https://www.declara.com/  Attitudes: the ways the learner thinks or feels about the To address this limitation of quantified-self devices in isolation, knowledge. This aspect includes personal, social and our vision is to take an educational perspective (associated with cultural views about the knowledge. A degree of trust is food literacy as a life-long learning problem) that would important in order for the learner to buy into an specific idea correspond to the first stage of our perspective of behavioural [13]. change (knowledge exploration). Then, we may include some self-tracking later stages, once an individual or various members  Self-efficacy: is defined as the confidence that individuals of the community can make sense and have better understanding have in their ability to plan and execute a course of action. of what they can look for in their own data. The key idea is to This aspect is closely linked with experience rather than generate the conditions to help them link the quantitative knowledge. The learner may be influenced by their sense of measures with higher level qualitative aspects of their own success/achievement, social models, persuasion by others, experience and literacy about food and nutrition. and their own personal agency [2]. In this paper, we focus on approaches to scaffolding knowledge, 4. PROPOSED APPROACH while bearing in mind that other elements such as attitudes and 4.1 Conceptual Proposal self-efficacy will need to be supported in order to scaffold Figure 1 presents a visual representation that situates the kind of sustained behavioural change in learners. tools we propose to provide support in the initial stages of behavioural change. A person’s journey towards making any 3.3 The Quantified-self for Health change in their behavior (in this case health or nutrition) begins The Quantified-self is a growing movement to incorporate with some initial attitudes (shaped by, for example, personal ubiquitous sensing technology into data acquisition on aspects of values, social pressure, culture, the resistance to change, food a person's daily life [27]. Considerable effort has been put on availability and security, etc) and environmental knowledge (what quantified-self solutions applied in health and wellness the learner knows about food, disease prevention, nutrients, improvement [8]. Different devices and trackers exist that information from the media, family and friends, etc) (see Figure automatically or semi-automatically keep records of goal 1–A). The aim is to provide a tool (B) that supports the first steps accomplishment, food consumption, portion sizes, physical for the learner towards the first outcome of the behavioural activity, caloric intake, sleep quality, posture, and other factors change (C): shaping and evolving new attitudes and knowledge that may affect individuals well-being. Some evidence has about the topic (food). At this point is where we define our reported that increased awareness about one’s own activity and approach as a Food Literacy problem. food consumption can motivate towards achieving personal goals (e.g. reach a certain body weight range), and that individuals can The subsequent steps of behavioural change are beyond the receive support from members of the community that also share purpose of the tool and this paper. It may involve the use of other their self-tracking experiences [6]. Thus, there has been some tools or mechanisms (D) to promote learner’s self-efficacy (E), enthusiasm about the key role that wearables can play in which then could lead to a sustained commitment (F). In these behavioural change strategies. However, as briefly described subsequent steps, it may be possible that self-tracking tools, food above, behavioural change requires a series of elements, from security programs (how to get better quality foods) and other knowledge exploration to self-efficacy, that are not necessarily community apps (e.g. community recipes) can provide a different supplied by a technological solution itself. Particularly, a recent type of support to develop self-efficacy, once the food literacy of study has reported how users show lower levels of engagement the learners had been improved. Building self- efficacy is needed, using smart wearables and self-tracking solutions as time passes particularly in this context, because each learner is in its own [21]. Thus, quantified-self applications may not be long-term journey; has a different level of food literacy development; has sustainable solutions for behavioural change without a richer and different food requirements; and requires the development of more complete perspective. Moreover, many tracking devices are certain level of personal agency for making a sustained built with closed architectures adding a possible lack of data commitment. validity and reliability that may affect attitudes towards those As a result, Figure 2 illustrates our conceptual approach. The flow measures. of information begins at the top of the figure, with users collaboratively curating information that they consider relevant to improve their overall health. For conducting user studies, we will need to identify specific populations (e.g. students of a university, young adults of certain age, etc) and, potential areas of interest or improvement for those individuals. Then, the second layer corresponds to the learner’s data that we can collect from user’s interactions with the curation tools. We will use the API provided by the tools aforementioned, to obtain user logs, which can be grouped in three categories: a) Figure 1. Situating the tools and awareness mechanisms in a wider view towards Curated information (which include behavioural change the text, videos and other resources that are extracted from the curated webpages); b) Collaborative metrics of resources viewed but not used, temporal learning evidence (which include the logs of user’s activity such sequences, durations of webpage views, and collaborative as comments added to others’ curated resources, discussion symmetry metrics [16]. Alternative analysis can also be threads, messages, etc.); and c) Information about the sources of done on the process to curate information by fore example, information (classifications of webpage types, links within the identifying the steps that most successful achievers follow. pages, and other meta data).  Knowledge sharing. Learning analytics can also provide The third layer shows the types of learning analytics outputs and with clues about how learners interact with other learners, techniques that we want to build by exploiting the learner’s data. share information, influence others or learn from interaction. In order to guide the design of the learning analytics needed, our Relevant forms of Social Learning Analytics [24] would approach will be grounded on the following guiding aspects for include social network analysis (e.g. based on social ties the development of knowledge related to learners’ food literacy. formed through peer discussion, and annotation of peers’ These aspects are: resources, to support the understanding of community structure and authority), discourse analytics (e.g. to provide  Knowledge curation. Learning analytics about the actual insight into the quality of argument in online interactions curation itself should include the generation of indicators [17], and writing analytics (e.g. to provide feedback to that may provide evidence about the curation process and learners on their reflections about how their efforts are information about the content that is being curated. Simple progressing [3]. semantic analysis on the curated text, such as topic extraction, aggregations of key terms could provide an  Knowledge certainty [13]. A key aspect for the learning overview of the learner’s curated content to recommend analytics tool to promote a shift in learners understanding is similar resources to the learner or highlight overseen topics. by providing the means to enhance their trust on the Learning indicators about the collaborative content curation information that is being curated. This is one of the major can include for example: webpage use metrics, bookmarks, problems indicated with current practices in health Figure 2. Conceptual approach: Making visible aspects of collaborative content curation for supporting food literacy in the initial stages of behavioural change information seeking highlighted in previous sections. wearable device. European Journal of Cultural Studies, 18, 4-5, 479- Learning analytics may help in providing a meta-analysis of 496. the sources that the learner has curated. For example, a [9] Estellés, E., Del Moral, E., and González, F. 2010. Social bookmarking tools as facilitators of learning and research collaborative processes: topology of sources could be defined to differentiate The Diigo case. Interdisciplinary Journal of E-Learning and Learning scientific from opinion-based sources of information. Objects, 6, 1, 175-191. Finally, the fourth layer corresponds to the means by which this [10] Fordyce-Voorham, S. (2015). A Food Literacy Model. Retrieved from http://www.foodskillsaustralia.com.au/wp- information will be mirrored back to the learners. This may content/uploads/2015/10/Food-Literacy-Model.pdf include simple dashboards with visualisations depicting the [11] Fox, S. 2006. Online health search. Pew Internet & American Life different learners’ data about the knowledge curation process (a Project. Retrieved from learning analytics approach). By contrast, the traces of the http://www.pewinternet.org/2006/10/29/online-health-search-2006/ learners’ activity could be aggregated into learner’s and/or social [12] Frawley, J. K., and Underwood, J. 2014. Designing Mobile Learning models. These can then be mirrored back to the learners using to Support Public Food Literacy: Constructivist Pedagogies and ICT metaphors and other simple visual aids (an open learner model Ecologies. Transactions on Mobile Learning, 10. approach), or be used to generate recommendations or suggest [13] Hsu, M.-H., Ju, T. L., Yen, C.-H., and Chang, C.-M. 2007. Knowledge sharing behavior in virtual communities: The relationship new unexplored content. between trust, self-efficacy, and outcome expectations. International Journal of Human-Computer Studies, 65, 2 (2//), 153-169. 5. FUTURE WORK DOI=http://dx.doi.org/10.1016/j.ijhcs.2006.09.003 This paper aims to trigger discussion about the type of data and [14] Kay, J., and Bull, S. 2015. New Opportunities with Open Learner presentation mechanisms that may help life-long learners set a Models and Visual Learning Analytics. In Proceedings of the stronger basis for behavioural change in the subsequent stages. In International Conference on Artificial Intelligence in Education particular, the application of our transdisciplinary approach to (Madrid, Spain June 22-26). 666-669. support food literacy and initial behavioural change requires [15] Kivits, J. 2009. Everyday health and the internet: a mediated health standing on three very different areas of research and perspective on health information seeking. Sociology of Health & Illness, 31, 5, 673-687. DOI=10.1111/j.1467-9566.2008.01153.x development: learning analytics, behavioural change, and food [16] Knight, S. 2015. Learning Indicators in SCIS Tasks. In Proceedings literacy. In general, the ideas proposed in this paper are centred on of the Proceedings of the 2015 Workshop on Evaluation on learning analytics for collaborative content curation in any Collaborative Information Retrieval and Seeking (Melbourne, context. For example, we plan to support students enrolled in our Australia, 2812378: ACM, 11-13. DOI=10.1145/2812376.2812378 Master of Data Science and Innovation program, who already [17] Knight, S., and Littleton, K. 2015. Discourse Centric Learning curate web resources as part of their natural learning practice Analytics: Mapping the Terrain. Journal of Learning Analytics, 2, 1, using Diigo. Alternatively, we may explore the potential of 185-209. learning analytics to enhance the awareness of students learning [18] Knight, S., and Littleton, K. 2015. Learning through collaborative information seeking Collaborative Information Seeking. Springer, how to curate resources to write up literature reviews in a research 101-116. methods subject. [19] Knight, S., Buckingham Shum, S., and Littleton, K. 2014. Epistemology, assessment, pedagogy: where learning meets analytics 6. REFERENCES in the middle space. Journal of Learning Analytics, 1, 2, 23-47. [1] Abraham, C., and Michie, S. 2008. A taxonomy of behavior change [20] Lambert, S. D., and Loiselle, C. G. 2007. Health information— techniques used in interventions. Health psychology, 27, 3, 379. seeking behavior. Qualitative health research, 17, 8, 1006-1019. [2] Bandura, A. 1977. Self-efficacy: toward a unifying theory of [21] Ledger, D., and McCafrey, D. (2014). How the Science of Human behavioral change. Psychological review, 84, 2, 191. Behavior Change Offers the Secret to Long-Term Engagement. [3] Buckingham Shum, S., Sándor, Á., Goldsmith, R., Wang, X., Bass, R., Endeavour Partners. Retrieved from and McWilliams, M. 2016. Reflecting on Reflective Writing http://endeavourpartners.net/white-papers/ Analytics: Assessment Challenges and Iterative Evaluation of a [22] Meinhold, J. L., and Malkus, A. J. 2005. Adolescent Environmental Prototype Tool. In Proceedings of the International Learning Behaviors Can Knowledge, Attitudes, and Self-Efficacy Make a Analytics & Knowledge Conference (Edinburgh, UK, April April 25 - Difference? Environment and behavior, 37, 4, 511-532. 29 2016). LAK '16. New York: ACM, to appear. [23] Rotman, D., Procita, K., Hansen, D., Sims Parr, C., and Preece, J. [4] Bull, S., and Kay, J. 2007. Student Models that Invite the Learner In: 2012. Supporting content curation communities: The case of the The SMILI Open Learner Modelling Framework. International Encyclopedia of Life. Journal of the American Society for Journal of Artificial Intelligence in Education, 17, 2, 89-120. Information Science and Technology, 63, 6, 1092-1107. DOI=http://iospress.metapress.com/content/d1363l5752666874 [24] Buckingham Shum, S., and Ferguson, R. 2012. Social Learning [5] Bunton, R., Murphy, S., and Bennett, P. 1991. Theories of behavioural Analytics. Journal of Educational Technology & Society, 15, 3. change and their use in health promotion: some neglected areas. [25] Siemens, G. 2012. Learning analytics: envisioning a research Health Education Research, 6, 2, 153-162. discipline and a domain of practice. In Proceedings of the [6] Chamberlain, A., Poole, E., Munson, S., Danis, C., and Churchill, E. International Conference on Learning Analytics and Knowledge 2015. Moving Beyond e-Health and the Quantified Self: The Role of (Vancouver, Canada, April 29 - May 2). LAK '12. New York: ACM, CSCW in Collaboration, Community and Practice for 4-8. DOI=10.1145/2330601.2330605 Technologically-Supported Proactive Health and Wellbeing. In [26] Siemens, G., and Baker, R. S. J. d. 2012. Learning analytics and Proceedings of the 18th ACM Conference Companion on Computer educational data mining: towards communication and collaboration. Supported Cooperative Work & Social Computing, ACM, 273-276. In Proceedings of the International Conference on Learning Analytics [7] Choudhury, M. D., Morris, M. R., and White, R. W. 2014. Seeking and and Knowledge (Vancouver, Canada, April 29 - May 2). LAK '12. sharing health information online: comparing search engines and New York: ACM, 252-254. DOI=10.1145/2330601.2330661 social media. In Proceedings of the Proceedings of the 32nd annual [27] Swan, M. 2013. The quantified self: Fundamental disruption in big ACM conference on Human factors in computing systems (Toronto, data science and biological discovery. Big Data, 1, 2, 85-99. Ontario, Canada, 2557214: ACM, 1365-1376. [28] Vidgen, H. A., and Gallegos, D. 2010. Food literacy: Time for a new DOI=10.1145/2556288.2557214 term or just another buzzword? Journal of the Home Economics [8] Crawford, K., Lingel, J., and Karppi, T. 2015. Our metrics, ourselves: Institute of Australia, 17, 2. A hundred years of self-tracking from the weight scale to the wrist