concept
active
concept:expected-free-energyExpected Free Energy
Free energy expected under future outcomes; guides policy selection via epistemic and extrinsic value.
Neighborhood — ranked by edge-count
Thinkers (1)
thinker
- Francesco Rigolistudies
Frameworks (1)
framework
- Active Inferenceassociated_withimplementsFoundational framework by Karl Friston; the paper extends it to three hierarchical levels for modeling meta-awareness.
Claims (1)
claim
- Key insight into structure of decision-making; explains intrinsic motivation and curiosity.
Methods (1)
method
- Softmax policy selectionimplementsSelecting policies using a softmax (normalized exponential) function of negative expected free energy.
Concepts (15)
concept
- Ambiguity (Expected Free Energy)associated_withrelated_toExpected entropy of outcomes given states; resolved by selecting states that yield unambiguous outcomes.
- Risk (Expected Free Energy)associated_withrelated_toKL divergence between predicted and preferred final states or outcomes.
- free energyrelated_toThermodynamic potential ΔF = ΔE − TΔS; domain walls form if ΔF < 0
- Action Selectionassociated_withimplementsChoice of policies minimizing expected free energy to realize preferred future states.
- Epistemic ExplorationimplementsBayes-optimal exploration driven by uncertainty minimization; natural behavior in active inference without handcrafted mechanisms.
- Extrinsic Valueassociated_withExpected evidence for preferred outcomes; utility-weighted term in expected free energy.
- Extrinsic Value / Pragmatic Valueassociated_withThe component of expected free energy that drives utility-maximizing actions based on prior preferences over outcomes.
- ambiguityassociated_withMultiple possible meanings for words like Alice, disambiguated by context; harder when grammar and meaning intertwine
- Epistemic Valueassociated_withExpected information gain about hidden states; drives curiosity and novelty-seeking; mutual information term in expected free energy.
- Expected Ambiguityassociated_withThe expected conditional entropy of outcomes given hidden states; lowering ambiguity favors states that solicit unambiguous observations.
- Riskassociated_withExpected complexity component of expected free energy; minimized by risk-sensitive behavior
- Noveltyassociated_withNew term in expected free energy representing information gain about the likelihood mapping; drives ignorance resolution
- Expected Costassociated_withThe KL divergence between predicted and preferred outcomes, minimized by policies that realize prior preferences.
- Exploration-Exploitation Trade-offassociated_withThe balance between gathering new information (exploration) and using existing knowledge to obtain rewards (exploitation).
- Ignoranceassociated_with
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Minimizing expected free energy for planning, decision-making, and action selection.
- Unifying nature of expected free energy claimed in Section 2 and 7.
- Physical quantity sharing same minimum as variational free energy (via Jarzynski equality); proxy for computational cost
- A foundational variational principle from statistical physics that formalizes how self-organizing systems maintain structural integrity and adapt to their environment by minimizing free energy—a mathematical bound on surprise or prediction error. Originally developed by Karl Friston, the framework unifies action, perception, and learning as processes of active inference, where systems both update internal models of the world and act upon it to reduce the divergence between predictions and observations.
- Definitional claim from Section 2.
- A key decomposition presented in Section 7.
- Expected log likelihood of data under posterior beliefs; measures fit to observations.
- Decision-making rule in active inference.