Definition
An intervention applied at the moment of decision that interrupts fast, heuristic acceptance of an AI output and pushes the person into analytical reasoning. Three tested designs: hiding the suggestion until it is requested, requiring the person to commit to their own answer first and imposing a delay.
What is at stake
It is the tested implementation of the third path, and it comes with its own objection. In Buçinca's study the designs that cut overreliance most were also trusted least, preferred least and rated most complex, and the authors say such systems may find less adoption.
Where the word came from
- Forcing function in designDonald Norman, The Design of Everyday Things, 1988, a constraint that makes the wrong action impossible
- Cognitive forcing strategies in diagnostic medicinePat Croskerry, 2003, who argued most diagnostic error is cognitive rather than knowledge-based
- Brought into interface design and measuredBuçinca, 2021
How episcope came to it
The term comes from outside the research base. The measurement does not. Buçinca's forcing functions study measured it: friction cut blind agreement on the AI's errors from 64 to 48 percent.
What to watch for
The term is not episcope's, and the eighteen-year lineage in emergency medicine is what makes it worth using rather than a weakness. The trade-off travels with it: the designs that recovered the most judgment were the ones people liked least.
Papers behind it
Buçinca, Z., Malaya, M. B., & Gajos, K. Z. (2021). To Trust or to Think: Cognitive Forcing Interventions Can Reduce Overreliance on AI in AI-assisted Decision-making. Proceedings of the ACM on Human-Computer Interaction, 5(CSCW1), article 188.