Directory · S
New here? Start with the premise →
Surprise
2 min read · 537 words
Surprise is the system encountering input that did not match its prediction — and the response is one of the more useful diagnostic signals available.
The hardware runs continuous prediction. What is about to happen, what the other operator is about to do, what the situation is going to produce. Most predictions are roughly accurate; the system is competent at modeling the immediate environment. Surprise occurs when prediction diverges from actual: the prediction said one thing, reality produced another, the system registers the gap.
The mechanism is functional. Surprise alerts the system that its model is inadequate in this specific domain, prompting the model to update. The operator who is regularly surprised in some domain has model that is poorly calibrated for that domain; the operator who is rarely surprised has either accurate model or has stopped attending to what would surprise them. Both are possible; the diagnostic for which is occurring matters.
The other application: surprise produces specific cognitive effects beyond the model-update function. The operator in surprise is briefly more receptive to information; the encoded patterns are momentarily disrupted, allowing input that would otherwise be filtered to register. This is why surprise is often part of effective communication, art, learning — the brief disruption of expected pattern produces the conditions where new material can land. The complete absence of surprise produces operator who is engaged with input but not actually processing it freshly; the input is being absorbed into existing patterns without producing the updates that surprise would have allowed.
From the chair: notice what is producing surprise in the operator’s life. Each surprise is information about where the operator’s model is incomplete or inaccurate. The other operator who behaved differently than expected. The situation that produced an outcome the operator hadn’t predicted. The own response that surprised the operator (suggesting the operator’s model of themselves had a gap). Each is invitation to update the model where the model was wrong.
The other application: deliberately seek out conditions that are likely to produce surprise. The conversations with operators outside the operator’s usual circle. The encounters with material that doesn’t fit current understanding. The experiences that the operator has not had before. Each provides input that current model cannot fully predict, with the surprise that follows producing the updates that staying within familiar conditions would not produce. The operator who never encounters surprise often runs with increasingly outdated model, with the model’s errors accumulating across time.
The other discipline: do not interpret all surprise as bad. Some surprise is delight — the system encountered something that was better than predicted, in ways that produce the felt response of pleasant surprise. Some surprise is information — the system’s model needed update. Some surprise is disturbing — the system encountered something the model was not prepared to handle. Each warrants its own response. The operator who treats all surprise as threat closes off the model-updating that surprise would have produced; the operator who treats all surprise as delight may miss when the surprise is reporting genuine difficulty. Calibrated reading of what surprise is reporting, in each case, produces more useful response than uniform reaction across all surprises.