Phase 2: Deliver therapy, do an RCT. How well do CBT and BA work with this population? (BA has been studied with adolescents, but in very small trials). And does matching teens to their preferred therapy and therapist improve engagement and outcomes?
We’re also curious about other things. Can you tinker with BA so the focus is on activities with “big chunks of hours”—asking not just what’s a one-off thing you want to do (like “go fishing”) but what can become a weekly routine, productively eating up six hours a week?
Because if that happens, we set sights on another potential win: reduced screentime. Seems like there would be a big benefit for a hobby that sticks (rock climbing, playing guitar), a part-time job, a structured sport with year-round practices and community (volleyball). As best we can tell, BA research has not gone deep here.
There will be technical issues to solve. One is how to balance our two goals: testing whether preference-matching improves outcomes while also maintaining randomized comparison groups to assess which therapy works better.
One solution is to randomize within each preference group—teens who prefer BA would be randomly assigned to receive either BA or CBT, along with a usual-care comparison group where appropriate.
Another technical challenge would be measuring whether BA or CBT reduces screentime: self-reports here are notoriously inaccurate, and screentime tends to be spread across multiple devices (phone, laptop, watch, TV).
Phase 3: Invite predictions about our experiment. We will do something else unusual: ask non-participating therapists and high school counselors to register their predictions. Which treatments do they believe will work best for which teens? Our hypothesis is simple: People who publicly commit to predictions are more likely to revise their beliefs when data arrive.
Think of it as a learning mechanism. After all, we estimate there are something like 250 million teen therapy sessions per year in the U.S. A few small, durable improvements in treatment could have a large total impact.
