Human delegation effects of predictability gains

Determine whether the predictability, adherence, grounded-accuracy, and step-level evidence gains produced by step-level SOP delivery via the Model Context Protocol change actual human delegation decisions through empirical user studies.

Background

The study evaluates technical proxies for trust in LLM-based operational-process automation, including process adherence, execution-path consistency, grounded task success, and structured step-level evidence. It does not directly measure whether human supervisors or organizational decision-makers become more willing to delegate work to agents when these properties improve.

The authors explicitly identify the relationship between these technical predictability gains and real human delegation behavior as unresolved and state that answering it requires empirical user studies. This extends the paper’s findings from infrastructure-level trust mechanisms to their practical effect on adoption and delegation decisions.

References

Whether these predictability gains shift actual human delegation decisions remains an open empirical question requiring user studies.

— One Step at a Time: Trading LLM Autonomy for Process Predictability  (2610.07817 - Schabert et al., 6 Oct 2026) in Limitations, item 7, “Trust is measured indirectly”