I dunno! In the paper, the link to the OSF is described as "containing more detailed information", but only has a vague description of the research question and it lists several questionnaires, roughly half of which do not get reported in the paper.

Yes, they did. I am considering emailing the editor to restate that I think this misrepresentation should be a serious enough violation to qualify for rejection.

This is in an Elsevier journal, by the way.

This way, she gets a time series of model-predicted attentional states for each participant. These states predict a variety of independent behavioral patterns, including speeding up of RTs under mind wandering, self-reported task-unrelated thoughts, and a shift towards these thoughts across time.

Second, Cathy @cathyzhang.bsky.social has developed a new framework to infer mind wandering (lapses of attention) from behavior. She combines the GLM-HMMs from @jpillowtime.bsky.social's group with a parametric version of the Sustained Attention to Response Task (SART).

osf.io/preprints/ps...

People trade off this cost with interference demands! They become more likely to choose lists of more incongruent Stroop trials if this minimizes switches between attentional states. This work nicely complements the work on "reconfiguration costs" by Ivan Grahek from the @shenhavlab.bsky.social.

Yet, the effect (I think) captures the system setting itself up for subsequent conflict (not current conflict). Thinking of this as a model-free (implicit?) form of adjustment for the future makes sense to me, because control demands are typically more predictable than in cog psych experiments.

I agree with Harrison: a bit of both. The adjustment is reactive: it occurs in response to demand. Moreover, it is observed in blocks with 50% congruent trials. In those, the congruency of the next trial can't be predicted: thinking of these adjustments as (intentionally) proactive feels silly.