Decision Intelligence

How can a system choose an action under uncertainty and show why that choice beats the alternatives?

Predictions are not decisions. A probability that a customer will pay does not say whether to call, wait or escalate, because the right action depends on costs and not on the probability alone.

Decision theory ties probability to cost. Choosing the action with the highest expected utility makes the trade-off explicit and lets the same model be reused when costs change.

a* is the chosen action, U the utility of action a, and the expectation is taken over the uncertain outcome given what is known, x.

1. Define the actions and the cost of each outcome.

2. Estimate outcome probabilities and check their calibration.

3. Compute expected utility per action and choose the maximum.

4. Show the decomposition, so a person can see which term drove the choice.

1. Evidence x

2. Outcome model P

3. Utility U

4. Expected utility per action

5. Choice

6. Explanation

No published experiments. The Agent Evaluation lab computes a cost-optimal confidence threshold on synthetic data, and the Decision Optimization lab chooses actions under a capacity limit.

Agent Evaluation · Decision Optimization

No original research results are claimed. This area is presented as a technical perspective with a computable demonstration on synthetic data.

• Expected utility needs costs that are hard to estimate.
• A miscalibrated probability makes the best action wrong.
• Fairness and regulatory limits belong in the constraints, not in the utility.

Background reading by other authors. These are not Menlo Tech publications, and no Menlo Tech paper is cited because none has been published.

• Berger, J. Statistical Decision Theory and Bayesian Analysis, 2nd ed. Springer, 1985.