Results for "alternative outcomes"

Tendency to trust automated suggestions even when incorrect; mitigated by UI design, training, and checks.

Measures divergence between true and predicted probability distributions.

Methods for breaking goals into steps; can be classical (A*, STRIPS) or LLM-driven with tool calls.

Measures how one probability distribution diverges from another.

Formal framework for sequential decision-making under uncertainty.

Number of steps considered in planning.

Sample mean converges to expected value.

Ensuring AI systems pursue intended human goals.

Maximizing reward without fulfilling real goal.

Model optimizes objectives misaligned with human values.

Model relies on irrelevant signals.

Probabilities do not reflect true correctness.

Ability to inspect and verify AI decisions.

Review process before deployment.

Continuous loop adjusting actions based on state feedback.

Combining simulation and real-world data.

RL using learned or known environment models.

Optimizing continuous action sequences.

Reward only given upon task completion.

Learned model of environment dynamics.

Learning by minimizing prediction error.

Imagined future trajectories.

AI systems assisting clinicians with diagnosis or treatment decisions.

AI that ranks patients by urgency.

Testing AI under actual clinical conditions.

Predicting case success probabilities.

Mechanics of price formation.

AI selecting next experiments.

Agents have opposing objectives.

Rules governing auctions.