Hypothesis evaluation based on ubicomp sensing: moving from researchers to users Farahnaz Yekeh School of Information Technologies The University of Sydney Sydney, NSW 2006, Australia fyek9388@uni.sydney.edu.au Abstract. Emerging pervasive sensing technology provides new ways to create persuasive systems that can help people improve their health. Much persuasive computing research has involved the exploration of re- searchers’ hypotheses about the ways that such ubicomp sensing can improve health. This thesis aims to enable individual users to test their personal hypotheses about how their actions, as tracked by ubicomp sen- sors, and the interface tools that they elect to use, actually impact their health goals. The key contributions are to develop the infrastructure and interfaces for a new persuasive system to support personal health hypothesis evaluation. Keywords: Persuasive technology; ubicomp sensing; personal hypoth- esis evaluation; personal informatics. 1 Introduction and Related Work A hypothesis is “an idea or explanation for something that is based on known facts but has not yet been proved”1 . It is also defined as “a proposition, or set of propositions, set forth as an explanation for the occurrence of some specified group of phenomena, either asserted merely as a provisional conjecture to guide investigation (working hypothesis) or accepted as highly probable in the light of established facts”2 . We define a personal health hypothesis as an individual’s belief about the ways that their actions affect their health. For example, a person may believe that if they significantly increase their level of physical activity, this will improve their health, in line with current health recommendations [4]. Other examples include: ‘If I eat less fat and oil, I will be healthier ’ and ‘If I restrict my consumption of carbohydrate, I will loose at least the rate of 2 kg in 8 weeks’. A large body of persuasive and ubicomp research has explored research hy- potheses related to improving or maintaining health. For example, the seminal work of Consolvo et al. [3] studied effects of UbiFit on the level of activity people maintained. UbiFit is an exemplar of a ubicomp sensor based system designed 1 http://dictionary.cambridge.org/dictionary/british/hypothesis?q=hypothesis 2 http://dictionary.reference.com/browse/hypothesis 2 Doctoral Consortium to improve health. It provided glanceable display on the user’s phone, enabling them to notice, or regularly check, their recent levels of activity. Essentially, that work evaluated the researchers’ hypothesis that people achieve and main- tain higher activity levels in the long term if they can readily see how active they have been recently. Li et al. [7] explored of various forms of personal data to improve awareness of physical activity. There has been considerable research evaluating various hypotheses about the ways that tracking individual’s data affects a behavior, an attitude or both (such as [3], [2], [6], [8]), but personal hy- pothesis evaluation, based on the individual user setting and testing their own hypotheses, has not yet been considered. 2 Thesis Contributions and Research Plan The aim is to provide a framework and the required interfaces, which make it possible for an individual to define, formulate and evaluate personal health hypotheses. We will study researchers’ hypotheses and identify a class of personal health hypotheses which individuals would like to evaluate. Then we will select the required data and ubicomp sensors to collect such data. A framework for this purpose wi ll be designed and implemented. This framework, will provide the possibility for individuals to manipulate data (add, edit, or delete), monitor, visualize, analyze and evaluate the results. Then we will evaluate the system in terms of usability of the interfaces, efficiency, performance and actual use. This thesis involves an iterative process with the following stages: 1. Design an architecture for personal health hypothesis evaluation. 2. Implement the architecture, based on personis. 3. Create interfaces for hypothesis associated with physical activity and health (based on creating a hypothesis framework associated with sensors) 4. Conduct usability studies to assess whether people can use the interfaces effectively and to gain insights into the ways people would like to explore personal hypotheses. 5. Field trial to learn about use. The key contribution is defining a new approach of personal informatics based on ubicomp sensing to achieve long term goals such as being healthier. We be- lieve that our approach may be valuable if evaluation of a personal hypothesis encourages individuals to be creative and out their own approaches to achieving their goals. For a personal hypothesis, we are concerned with one participant, the individual person who wants to test a hypothesis. Therefore, despite the broader definition of a hypothesis, we will not deal with population level outcomes. 3 Preliminary Results The aim is to create an opportunity for a person to take a new form of control over the ways that they make use of ubicomp sensor technology to help achieve their health goals. The primary results are defining elements of a personal health hypothesis and a proposed architecture, which are explained in this section. Doctoral Consortium 3 3.1 Elements of a personal hypothesis As a foundation for defining our approach, we introduce an illustrative scenario. Alice has just been diagnosed with mild hypertension (high blood pressure). She decides she wants to try altering aspects of her lifestyle to tackle this. She has been given an overwhelming amount of literature and advice. This points to several possible hypotheses about reducing hypertension. She decides she would like to test the hypothesis: if I exercise more that will reduce my blood pressure. Alice might buy an activity sensor device3 to track her activity levels, and a device to measure her blood pressure4 . These ubicomp sensors could collect the data required to evaluate her hypothesis in the scenario. We now can define the elements of a personal wellness hypothesis that can be assessed by a ubicomp sensing system. 1. The user must formulate a suitable hypothesis. It takes the form if I do X then I will see effect Y over time period Z. 2. X and Y can be measurable by ubicomp sensors (potentially multiple sensors for each). 3. The user must establish the baseline values for X and Y. 4. The user must set appropriate goal values for X and Y. 5. The user must make use of the devices to measure X and Y over time period Z, making use of interfaces that are based on best practice for persuasive technologies. We now illustrate this in terms of our scenario. Suppose Alice sets the hy- pothesis: If I increase my level of activity, my blood pressure will drop. Armed with her FitBit, she can easily collect long term data that measures at least some aspects of her activity. Her starting point for assessing her hypothesis is to establish her baselines for X (activity level) and Y (blood pressure). To get the baseline, she should use the FitBit for a period of time, such as a week. For example, this may indicate she typically walks 5,000 steps a day on work days and 15,000 a day on weekends. Similarly, she can use the Withings device to measure her blood pressure. This also needs to be done over a period of time. This is because a single Blood Pressure reading is not a reliable indication of her true blood pressure. So she may take the early morning and evening measures each day for a week [9]. She then needs to decide a suitable time period to as- sess her hypothesis (Z). For example, she realizes that it may take several weeks for a change in activity to affect her blood pressure. So, she may select Z as 6 months. We note that persuasive literature indicates that she will benefit from feedback through that period [3]Such support can be readily provided by a ubi- comp based system. At the end of the period, she will need interfaces that help her assess whether her hypothesis was supported by the evidence. This means that she needs to be able to see if she did indeed increase her level of activity 3 for example, a FitBit http://www.fitbit.com 4 for example, a Withings Blood Pressure Monitor http://www.withings.com/en/bloodpressuremonitor 4 Doctoral Consortium and maintain that increase during the 6 month period. She also needs to get a measure of her blood pressure at that time, following a similar process to that used to determine the baseline. Although we illustrated our definition in terms of our scenario, personal hy- pothesis evaluation can be applied much more broadly. For example, it could make use of other sensors, even for this hypothesis. There are also many other ubicomp sensors5 that could support other hypothesis involving, for example, weight, intensity of activity and glucose response. There is also considerable scope for people explore hypotheses about other aspects of their lives, such as altering behaviors to reduce carbon footprint. 3.2 Proposed architecture The proposed architecture for hypothesis evaluation is shown in Fig. 1. A set of data based on the ubicomp sensing would be collected via the Sensors. Users con- trol the system using Hypothesis Setting, Sensor Linking and Hypothesis Evalu- ation interfaces. Hypothesis setting is based on the form ‘if I do X then I will see effect Y over time period Z’. The three parameters of X, Y and Z need to be set using the hypothesis setting interface (Fig. 1. b) and the values of X and Y will be collected over time period Z. For example, the parameters in our scenario’s hypothesis are activity level (X), blood pressure (Y) and 6 months (Z). To create user models and hypothesis, personis [1] will be used. This gener- alised framework has the power, flexibility and low cost for implementation and supports privacy and scrutability (means that users know which information are personalised for them and how the system decides to select them) [1], [5]. Fig. 1. a) Proposed Architecture, and b) Hypothesis Setting. 5 For example, Withings Body Scale http://www.withings.com/en/bodyscale to mea- sure weight and fat mass, Basis https://mybasis.com/ to track heartbeats, Body- Media http://www.bodymedia.com/ to track calorie burned. Doctoral Consortium 5 4 Conclusions and Next Steps The goal of this research is defining a new approach of personal informatics based on ubicomp sensing to achieve long term goals such as being healthier. An architecture is proposed to support personal health hypothesis evaluation. The next steps are to implement this infrastructure and assess how people use it. References 1. M. Assad, D.J. Carmichael, J. Kay, and B. Kummerfeld. Personisad: distributed, active, scrutable model framework for context-aware services. In Proceedings of the 5th international conference on Pervasive computing, Pervasive’07, pages 55–72, Berlin, Heidelberg, 2007. Springer-Verlag. 2. S. Chatterjee, Jongbok Byun, Akshay Pottathil, MilesN. Moore, Kaushik Dutta, and Harry(Qi) Xie. Persuasive sensing: A novel in-home monitoring technology to assist elderly adult diabetic patients. In Magnus Bang and EvaL. Ragnemalm, editors, Persuasive Technology. Design for Health and Safety, volume 7284 of Lecture Notes in Computer Science, pages 31–42. Springer Berlin Heidelberg, 2012. 3. S. Consolvo, P. Klasnja, D.W. McDonald, D. Avrahami, J. Froehlich, L. LeGrand, R. Libby, K. Mosher, and J.A. Landay. Flowers or a robot army?: encouraging awareness & activity with personal, mobile displays. In Proceedings of the 10th international conference on Ubiquitous computing, UbiComp ’08, pages 54–63, New York, NY, USA, 2008. ACM. 4. W.L. Haskell, I. Lee, R.R. Pate, K.E. Powell, S.N. Blair, B.A. Franklin, C.A. Macera, G.W. Heath, P.D. Thompson, A. Bauman, et al. Physical activity and public health: updated recommendation for adults from the american college of sports medicine and the american heart association. Medicine and science in sports and exercise, 39(8):1423, 2007. 5. J. Kay and B. Kummerfeld. Creating personalised systems that people can scrutinise and control: drivers, principles and experience. ACM Transactions on Interactive Intelligent Systems (TiiS), 2012. 6. S. Lederer, J. Mankoff, A.K. Dey, and C. Beckmann. Managing personal information disclosure in ubiquitous computing environments. 2003. 7. I. Li, A. Dey, and J. Forlizzi. Position paper on using contextual information to improve awareness of physical activity. First International Forum on the Application and Management of Personal Electronic Information, 2009. 8. I. Li, A.K. Dey, and J. Forlizzi. Understanding my data, myself: supporting self- reflection with ubicomp technologies. In Proceedings of the 13th international con- ference on Ubiquitous computing, UbiComp ’11, pages 405–414, New York, NY, USA, 2011. ACM. 9. T.G. Pickering, J.E. Hall, L.J. Appel, B.E. Falkner, J.Graves, M.N. Hill, D.W. Jones, T. Kurtz, S.G. Sheps, and E.J. Roccella. Recommendations for Blood Pres- sure Measurement in Humans and Experimental Animals Part 1: Blood Pressure Measurement in Humans: A Statement for Professionals From the Subcommittee of Professional and Public Education of the American Heart Association Council on High Blood Pressure Research. Circulation, 111:697–716, 2005.