Summary Response Week 11: Behavior Change & Personal Informatics

CPSC 601.25 Week 11 Part 2

Papers

In this response I will discuss a paper which reviews strategies for behaviour change from the medical literature and suggests some design guidelines for building 'behaviour change' applications.

Behaviour

Convincing patients to eat better, exercise more and take their medication has been a standing problem in medicine for years. Long term longitudinal studies on medications have to consider 'non-compliance' because completely rational people do not take their medicine. This is an interesting problem from the perspective of psychology but it's also a very practical problem for patients and doctors. Evidently in the medical field some evaluation has been done on 'interventions' to try to convince patients to change their behaviours. The researchers review some meta-analysis about this, as well as existing papers and books to extract strategies. They then consider how these strategies could be integrated in behaviour changing software.

The authors suggest 5 strategies, of which the first 3 relate specifically to so called 'action plans'. These action plans should be made up of goals which are short term (1) and actionable (2) - meaning that they are actions that people can take - and which the patient feels confident (3) in accomplishing. These are interesting, and consistent with other topics about self regulation, but seem sort of limited to an app whose job is to create these action plans. But it's possible that all behaviour changing apps will require this component. Another difficulty is how to programatically decide whether the proposed goals meet these conditions. One approach might be to offer specific and strict templates with sanity checks - allowing users to lose x number of pounds a week, etc. It's also difficult to asses what someone feels confident with, the authors suggest asking them on a scale of 1 to 10, but this is super lame. Another approach might be to target a goal that is below average and adjust it incrementally based on performance.

Constraints

The authors never give raw statistics on how effective these interventions generally are, but we can assume they are not very effective. If they were effective we would see less people dying of preventable choices that they've made, such as smoking, being overweight or sedentary, etc. From the field of psychology some concepts from self regulation could be coupled with these strategies. The Beeminder technology charges you money, if you deviate from your assigned quantifiable plan. Other apps try to add a social component to shame people who deviate from their stated goals.

Overall

This paper was surprisingly good for being a meta-analysis paper and being published in CHI. The paper presented clear design goals with examples and rarely descended into obscure and pointless language. It also seemed possible that they intended to develop software around these design goals and apply it to a real world problem.