This week’s blog post is about different methods we have tried in creating tailored messages for intervention studies. Soon healthcare professionals will probably be able to send you automatic reminders generated by ChatGPT, but whether such messages are actually helpful is a question for research. The use of a relevant behavior-change theory is likely to improve their effectiveness. https://sites.google.com/view/two-minds/blog
Let see, in general, you believe the best way to treat a human being with needs is to “tailor” (AI, ChatGPT, etc etc) advertising to get them to effectively “buy” what you have. ---- My suggestion; spend a year with someone who has actual healthcare needs to understand and experience what's it like navigating virtual systems. You may find helping real-live-people is different than "behavior-change theory".
Paul F Cook --- Paul, I have an additional comment. Yesterday ChatGPT was used to publish a photo of Trump being arrested. In my city it was also used to contact all people that are enrolled in SNAP, the old food stamps. Each was a perfect rendering. The fake mail was an exact rendering of the government assistance program. - Today I received video of Trump and Santos embracing one another. These experiences are leading me to be extremely wary. I wouldn't respond to any health need that wasn't a direct link to a health professional I know. And knowing how easily the entire health web site I use and its direct message to my doctor can be faked by ChatGPT I will be hesitant to rely on that........Again I'm sorry, but ChatGPT doesn't add to or sustain my most basic need for honesty & trust.
Hi, Lee. I tend to agree with you about the risks of ChatGPT and its brethren. My main argument in the blog is that we need tools that provide more careful responses, not just plausible messages based on predictive text modeling. It's concerning that these new tools deliver messages that seem reasonable at first glance but might actually be unhelpful or even false. So your second response is something that I'm concerned about also.
I'm less sure that I agree about the benefits of live human interaction as suggested in your first response. First, healthcare systems are already deep in the world of automated messaging: see this recent review in JAMA - Article Evaluating the Role of System-Generated Communications in He...
Regardless of whether it's the most desirable strategy, health care systems are going to continue and expand their use of these tools because of their low cost and high reliability. Second, although I think it is preferable to have a good interaction with a well-qualified human on a good day, I doubt that these conditions are always met in people's interactions with the health care system. Robots don't have bad days, get distracted, hate their job, or have other human flaws. (They have different flaws, robotic ones). In some cases, the tradeoffs might mean that robot interactions are actually more helpful.Overall, I think it's inevitable that automated systems are going to be a bigger part of healthcare in the future, just as they are already in customer service and other industries. The question for me, therefore, is how to do this well. I do agree that the automated interaction has to feel appropriate to the end user, the patient, if it's going to be effective. And I think it also has to be based in what we know from the study of behavior and what helps actual people to make actual changes in their health.
Paul F Cook .....Paul, below is link to a text file. It a long discussion of what I think, some examples and recommendations. Hope this helps.
- Badges
- Science topic