Long-term persistence, transfer, and generalizability of Engage-to-Unlock effects

Determine whether the effects of the Engage-to-Unlock mechanism persist over time, transfer to later unassisted performance, and generalize beyond writing tasks.

Background

The paper evaluates Engage-to-Unlock in a short-term argumentative-writing experiment involving 398 participants. The mechanism delays direct generative assistance until participants have articulated a position and supporting argument, and the experiment finds changes in effort allocation and downstream evaluation speed.

The authors explicitly state that the durability and scope of these effects are unresolved. In particular, it is unknown whether the observed benefits extend to later performance without AI assistance, remain after repeated use, or apply to domains other than writing. The paper recommends longitudinal studies examining learning, skill retention, and transfer across repeated use and other domains.

References

Although Engage-to-Unlock changed engagement and downstream evaluation, it remains unclear whether these effects persist, transfer to later unassisted performance, or generalize beyond writing.

— Who Thinks First? Designing Productive Friction with Engage-to-Unlock GenAI  (2610.01518 - Su et al., 1 Oct 2026) in Section 5, “Limitations and Future Work”