Facebook chatbot for older adults
Aiding older adults perform rapid dietary recall
Build a mobile experience sampling tool to support older adults in dietary recall
HYPO-BOT, a Facebook chatbot that aids older adults perform rapid dietary recall
HYPO-BOT is currently being tested on top of traditional ASA24 questionnaire in an exploratory study
Clinical research studies typically use the automated self-administered 24-hour dietary recall (ASA24) for studying diet and disease associations. Although extensive and thorough, ASA24 completion is somewhat time-consuming. Specifically, people who have low numeracy and literacy skills have great difficulty completing the tool on their own. This makes the ASA24 an expensive and less practical solution as a dietary information system for patient or provider feedback in disease prevention and care. Hence, there is a need for a system that allows a respondent to independently log food and drink consumption for low cost and with minimal effort.
Mobile experience sampling method (ESM) can provide a possible solution. With advancing technology, ESM has been executed on mobile devices such as smartphones, specifically, as text messages delivered via a smartphone application. Importantly, adults of most ages now have mobile devices capable of receiving text messages, and even smartphones are increasingly the norm, including in minority and lower socioeconomic populations. Although ESM is considered the gold standard of experiential sampling in health research, self-report through ESM has repeatedly been considered a burden given the need to administer instruments multiple times in a day. Further, the collected data can suffer from poor adherence and misreporting, especially if the instrument is cumbersome to use or doesn’t suit individually variable reporting needs and preferences (e.g., sleep and work schedules). This motivates the need for experience sampling tools that not only support a respondents’ needs, but also allows real-time self-report multiple times in a day with low burden while being easy enough to promote recurrent use.
The client's desire was to design a tool for those most vulnerable and critical to advances in chronic disease management. To narrow project scope, we focused on older adults of a safety-net health system who are taking medications for diabetes.
How it works
HYPO-BOT is a conversation system that reminds a user to self-report food or drink consumption in a Facebook (FB) private message. Some private messages pushed by HYPO-BOT include questions that elicit the name of food or drink, ingredient names, quantity for select ingredients, and if the food or drink was store or restaurant-bought. HYPO-BOT uses natural language processing (DialogFlow) to process a users’ open-ended or selected response and follow-up with a logical question. HYPO-BOT also includes a learning component that stores responses frequently reported by users. The stored information is used during future conversations to present context and aid rapid response.
Semi-structured interviews and card sorting to understand the mental model of older adults specific to food groups and common food items they consumed on a daily basis
Co-design with sample of target users using invisible design
Consultation with experts or stakeholders (including social scientist, exercise physiologist, and health services research scientist) at various points in the design process
Dietary recall using a simulated conversation agent in a controlled setting
Dietary recall using experience sampling in the field over 2-week period
Older adults can effectively recall food or drink consumption using a structured dialog with abstract images or icons
Follow a question with a set of response options to choose from
Use pictures of food/drink as abstract icons to aid rapid recall and selection
Older adults do not prefer using FB messenger’s keyboard because it has too many buttons and they are oftentimes spelling words incorrect
Maximize quick reply and webview options
Introduce spell-check for instances keyboard is used
Older adults want means to correct an action/error during a recall
Support exiting and re-starting a conversation
Support recursive reporting
Strategized and led product design research methods
Mentored and collaborated with design team to run invisible design workshops, create design probes, prototypes, and test simulated conversation system
Worked with research team to setup the bot for field trial