Talk:Evidence-based assessment/Back to the Future

ELIZA and Machine Learning

Here's a really neat comment on the prospect of technology changing therapy:

I was struck by the emphasis on automation and machine learning paired with big data as the way to the future. It almost seemed as if emotion detection software combined with automated empirically supported intervention could result in the extinction of the therapist (or clinical assessor) as we know it. As I was contemplating my eventual obsolescence, I realized there was a critical gap in the proposition which allowed for the clinician to retain a toe hold in the process of EBA (or maybe not). Nowhere in this well-written and provocative paper, do the authors consider the fact that a non-trivial percentage of our patients do not, either deliberately or unconsciously, provide accurate information regarding their clinical condition and emotional state. For example, self-report of alcohol or drug consumption does not correspond well with urine toxicology findings especially where there are incentives to underreport. Further, in VA settings there are strong secondary gain pressures where reported PTSD symptoms can lead to disability payment and perhaps adjudication of criminal charges through veterans courts. What is the role of symptom validity assessment in the proposed hybrid integrative model for EBA? After all, on the individual level garbage in is garbage out for any big data driven machine learning assessment model.

There is a lot to unpack here. For now, I will just start by saying that I agree. The Garbage In-Garbage Out (GIGO) problem will definitely hamstring a lot of early applications of machine learning, in particular. In addition to the points raised above, there are systemic issues with reimbursement that bias choice of billing diagnoses, which will further handicap initial implementations of machine learning.

But I also think that the larger trend is still that these methods are going to become increasingly common, and the setbacks will be speedbumps. I think that we will get to re-think what skills and qualities human beings add to the mix, and our role will change accordingly. Using a calculator or smartphone app to divide the check or figure out the tip doesn't mean that we'll stop dining out, just that we'll handle some aspects of it more efficiently.

More later, too -- this comment (and issue) warrant revisiting. Eyoungstrom (discuss • contribs) 14:58, 28 March 2017 (UTC)