The architecture functioned. Inclusion widened. Default rates stabilised.

Then behaviour began to bend.

Users learned which signals the system rewarded.

Stability metrics improved.

Predictability increased. Risk tolerance narrowed.

Not dishonesty optimisation.

Trust scores rose. Initiative softened.

The distortion was subtle: the system rewarded reliability, and people optimised for legibility. Necessary pivots felt penalised. Creative risk appeared inefficient.

When trust becomes measurable, behaviour adapts to measurement.

Programmable trust introduced a tension fintech had not fully anticipated: how to quantify reliability without compressing possibility.

The models were accurate. The human response was strategic.

And strategy, when sustained, reshapes culture.