Forcing Function
A forcing function is any structural mechanism that makes a particular action mandatory at a particular moment — interrupting the default flow of work to insert an explicit check. In medicine, the surgical timeout is a forcing function: before the first incision, the team must pause to verify patient identity, procedure, and site. The pause is not optional. It is built into the workflow such that proceeding without it requires explicit deviation rather than implicit omission. Gawande's work identified forcing functions as the mechanism through which checklists, timeouts, and structured handoffs produce improvement — not by reminding practitioners of what they already know but by making the pause structurally mandatory rather than motivationally optional.
In the AI Story
The concept comes from interaction design — Don Norman's 1988 The Design of Everyday Things identified forcing functions as constraints that prevent users from proceeding until a particular condition is met (the ATM that will not release the transaction receipt until the card has been retrieved). Medical quality improvement imported the term to describe workflow structures that perform the same function in professional practice.
The distinction between forcing function and reminder is the distinction between structural and motivational intervention. A reminder depends on the practitioner noticing it, choosing to act on it, and remembering the information it contains. A forcing function depends on none of these. It operates at the workflow level, making the desired action the path of least resistance rather than requiring the practitioner to generate the action against the grain of pressure, fatigue, or competing demands.
AI-assisted building has almost no forcing functions. Verification is motivational — builders verify when they choose to, with the thoroughness they choose. Under velocity and attentional narrowing, the choice systematically tilts toward under-verification. The institutional remedy is not exhortation but workflow redesign — building the pauses into the tooling, the deployment pipeline, the code review process, such that proceeding without verification requires explicit deviation rather than implicit omission.
Examples of AI-era forcing functions might include: pre-commit hooks that require explicit acknowledgment of AI-generated code segments; deployment pipelines that block until specified verification checks are completed; code review templates that structure reviewer attention to the categories of AI failure most likely to matter; pair-programming conventions that require articulation of the AI's architectural choices before acceptance. Each converts a motivational choice into a structural requirement.
Origin
The term originated in Donald Norman's 1988 The Design of Everyday Things (originally titled The Psychology of Everyday Things), which categorized forcing functions into interlocks, lock-ins, and lock-outs. The medical quality improvement literature adopted the term through the patient safety movement of the 1990s, particularly in the work of James Reason on human error and error-prevention architecture.
Gawande's treatment generalizes the concept across domains — surgery, aviation, construction, finance — and establishes it as the structural complement to the checklist methodology.
Key Ideas
Structural, not motivational. Forcing functions operate at the workflow level, independent of practitioner vigilance.
The path of least resistance. Correct action becomes the default, deviation requires explicit choice.
AI-era analogs. Pre-commit hooks, pipeline gates, review templates, pair-programming conventions can perform the function.
Exhortation is not sufficient. Telling builders to verify produces unreliable compliance; restructuring the workflow produces reliable compliance.
The design task. Identifying the decision points where forcing functions would catch high-consequence failures is itself an institutional discipline.
Appears in the Orange Pill Cycle
Further reading
- Donald Norman, The Design of Everyday Things (Basic Books, 1988/2013)
- James Reason, Managing the Risks of Organizational Accidents (Ashgate, 1997)
- Atul Gawande, The Checklist Manifesto (Metropolitan Books, 2009)
- Sidney Dekker, The Field Guide to Understanding Human Error (Ashgate, 2014)