Skip to main content

Designing Consumer Health Informatics Applications by Adrienne Pichon

Speaker's interests are interdisciplinary: Informatics, Design, Qualitative.

Theoretical Foundations for Design​

Informatics is not (cannot) be an incestuous field: you need to be curious about a lot of other fields and do your due research. Emphasize humility. You don't always (and most likely won't) do "ground-breaking" work. Understand and build upon others' work.

There is no 'a-political' when it comes to design. Either explicitly and implicitly (mostly the latter), you are embedding bias. (Think of Timnit Gebru's work at Google.)

The field borrows heavily from health psychology and behavioural science. "Behaviour changes theories". A lot more in Prof. Mamykina's Research Methods class.

  • There's a Health Benefit Model which appears to be a simple incentive → outcome setup. Don't have access to transit? Well, it's lower likelihood that you'll get the flu shot. Some others:
  • Transtheoretical Model: Precontemplation → Contemplation → Preparation → Action → Maintenance. It deals with Stages of Change. How much further are you along in this chain?
  • Fogg Behaviour Model: Nice decreasing slope with Ability on X-axis and Motivation on Y-axis. There's an "Action Line" that delineates where prompts fail and succeed (below and above respectively). It's simple but it has problems and ones you can guess. You may have the motivation but no bus change.
  • Self-Determination Theory: Competence + Relatedness + Autonomy. This is related to internal motivation being more durable than external.

HCI and Informatics Models​

  • Personal Informatics Models: How people Track → Reflect → Act on data. Full flow is Prep → Collect → Integrate → Reflect → Act. Note that this is iterative. There's a "Lived Model" that's really more reflective of the complications of being human. You start and stop and resume.
  • Sensemaking Models: How people interpret complex information. Read about Mann Gulch Disaster. It's about ingesting, integrating information and not just collecting it. Weave stories to get people to act. If not, you just have a nice Relational Database.
  • Technology Acceptance Models: Wheter or not people will actually use what you build. What prevents people from adopting your shiny and helpful new thing?

Human-Centered AI​

Thing is: there are many ideas but not much of a consensus on what it is: How do we, what does it mean to center a human in your efforts?

Application Paradigms and System Design​

There are sevel paradigms for HCI applications. There was an early 'Personal Informatics' movement (see book called Quantified Self).

Next, there's a lot of emphasis on Visualization which carries a lot of the aims here. It's never about making things pretty: how are you helping people make decisions and achieve their goals?

Then there's ye olde RecSys. They're fun but just remember that that the stakes are always higher in healthcare.

A different idea on explainability... TODO ask later.

There's Nudging which you've encountered when paying tips. Think of how you'd frame calls to action.

Now Coaching aims to help people make decisions on their own. TODO: Is this a 'superset' of Nudging? (No this gives progress reports, feedback)

There are tools you'd design for Reflection and Sensemaking. E.g. How can you use tech to go over your journals or a simple daily mood-o-meter and see trends and generate meaning you wouldn't see otherwise?

Games! Gamifying things would help motivate you do things. E.g. Duolingo, the circles on the Apple Watch, etc. Thing is it has to connect to people's personal goals. Simply saying "You beat X" or "filled this circle" won't keep people engaged for too long.

Just-In-Time Adaptive Interventions (JITAIs). Trying to quit drinking? This would act on your 'state of vulnerability' (e.g. GPS location, time of day) and buzz your wrist to keep driving as you approach a liquor mart.

Finally, there's just your bread-n-butter ML/AI models.

Empowerment​

Patient Empowerment is the ultimate goal of HCI. You can do this with visualizations and dashboards, relection and sensemaking interfaces, nudge-n-coach systems, gamification.

TODO: What do you do when you deal with people who come from cultural settings where they cede authority to caregivers? (Talk to people, patient and clinician, Center the Human... many times the answer is to not do anything).

There's a whole scale of "Shared Decision Making" which has proportions of responsibility shared between patient and physician.

TODO: Is this a 1-1 setting? What if I go to my physician and think that Vaccines are evil and my decision has an impact on others? (When there's a community component, designing and solving problems definitely goes beyond 100% patient autonomy.)

Discussion on EHR Integration challenges: these barriers are technical and organizational.

There are 'macro' Population and Research Applications too (like Digital Phenotyping).

Critical Perspectives and Future Directions​

Again, you do not design or use or deploy any tech without ethical, legal, social, psychological concerns. These are inextricable parts of your efforts (someone from Meta should've sat in).

So be a good person and think very carefully about Privacy, Consent, Security, Bias, Equity, Justice, Autonomy. These are as important as as your spiffy new interface backed by an amazing model.

Multimodal sensing. LLMs and GenAI. Closed-Loop Systems (e.g. Insulin pumps). Things need to be personally grounded and attuned to peoples' lived experience. Be kind.