Uploaded July 2026 | Updated November 2026, 16 hours ago
"Understanding Integrative Complexity Through Active Inference"
Alexander Hemming, Dylan Grove
Recent studies have shown a link between integrative complexity (IC) and neuroscience, helping understand how the mind addresses varying input types into a cohesive model. Likewise, the application of IC scores to uncertainty-related tasks have yielded results linking lower ambiguity and deliberation with higher IC scores, linking complexity to gaining model evidence thus reducing personal uncertainty. In this paper we provide a computational account of IC from the active inference framework (AIF), linking the process of structured learning with the cognitive function of IC. Agents maintain and update beliefs over competing world states and subsequently integrate them progressively into richer (or, more complex) generative models through structure learning. This process bears strong resemblance to IC as a measure of reasoning, specifically, the differentiation of competing hypotheses and their integration into coherent, higher-order understanding. IC can therefore be understood as a cognitive-level observable of the underlying AIF process, with AIF in turn providing IC with a computational and neurobiological mechanism for explaining how humans construct models of increasing complexity over differing timescales. We establish that IC accounts alone are strengthened by this updated framework which expands its applicability and grounding into baseline cognitive mechanisms. We therefore show two kinds of IC, dialectic and elaborative, reinterpreted through active inference frameworks as reasonable proxies of abstract depth within a generative model. This provides a meaningful bridge for active inference to have a window into political and linguistic theory, with implications for how active inference scholars approach sociopolitical models, alongside political theoretic contributions to discourse and deliberative democracy.
----
Active Inference Institute information:
Website: https://www.activeinference.institute/
Activities: https://activities.activeinference.institute/
Discord: https://discord.activeinference.institute/
Donate: http://donate.activeinference.institute/
YouTube: youtube.com/c/ActiveInference
X: https://x.com/InferenceActive
Active Inference Livestreams: https://video.activeinference.institute/
------
CHAPTERS
0:00 Welcome and introductions
3:00 Integrative and cognitive complexity
4:30 Central prediction: open, pluralistic modeling
6:00 Measuring integrative complexity
7:30 Differentiation across domains
9:00 The scoring scale
12:00 Hierarchical integration
13:30 Cognitive structure and dialectical tension
15:00 Scoring participants
16:30 Performative complexity in elite decision-making
18:00 Automated scoring (AutoIC)
19:30 Precision allocation
21:00 Two time scales and temporal depth
22:30 Integrative complexity as abstract depth
24:00 Differentiation as competing hypotheses
25:30 Integration as structure learning
28:30 Language as an expression of the model
31:30 Responding to new problems
33:00 Model reduction and consolidation
34:30 Multi-agent settings and performativity
36:00 Elaborative complexity: a converged model
40:30 Dialectical complexity: holding alternatives
45:00 Understanding contradictory views
46:30 Elaborative vs. dialectical profiles
49:30 Task designs
51:00 Counterfactual deliberation
52:30 Complexity as state-dependent precision
54:00 Extremism and converged models
57:00 Encouraging both forms of complexity
1:01:30 Limitations and empirical tests
1:04:30 Discussion
1:09:00 Discussion: democracies over long time scales
1:13:30 Discussion: agents that model each other
1:16:30 Discussion: structure learning and precision
1:21:00 Discussion: automatic scoring and model fitting
1:24:00 Discussion: gradient-free political landscapes
1:27:00 Discussion: Dirichlet counts and coarse-grained hypotheses
1:33:00 Discussion: context and changing worlds
1:36:00 Discussion: voting and multi-agent games
1:40:30 Discussion: context-sensitive preferences
1:43:30 Discussion: neuromodulators
"Understanding Integrative Complexity Through Active Inference"
Alexander Hemming, Dylan Grove
Recent studies have shown a link between integrative complexity (IC) and neuroscience, helping understand how the mind addresses varying input types into a cohesive model. Likewise, the application of IC scores to uncertainty-related tasks have yielded results linking lower ambiguity and deliberation with higher IC scores, linking complexity to gaining model evidence thus reducing personal uncertainty. In this paper we provide a computational account of IC from the active inference framework (AIF), linking the process of structured learning with the cognitive function of IC. Agents maintain and update beliefs over competing world states and subsequently integrate them progressively into richer (or, more complex) generative models through structure learning. This process bears strong resemblance to IC as a measure of reasoning, specifically, the differentiation of competing hypotheses and their integration into coherent, higher-order understanding. IC can therefore be understood as a cognitive-level observable of the underlying AIF process, with AIF in turn providing IC with a computational and neurobiological mechanism for explaining how humans construct models of increasing complexity over differing timescales. We establish that IC accounts alone are strengthened by this updated framework which expands its applicability and grounding into baseline cognitive mechanisms. We therefore show two kinds of IC, dialectic and elaborative, reinterpreted through active inference frameworks as reasonable proxies of abstract depth within a generative model. This provides a meaningful bridge for active inference to have a window into political and linguistic theory, with implications for how active inference scholars approach sociopolitical models, alongside political theoretic contributions to discourse and deliberative democracy.
----
Active Inference Institute information:
Website: https://www.activeinference.institute/
Activities: https://activities.activeinference.institute/
Discord: https://discord.activeinference.institute/
Donate: http://donate.activeinference.institute/
YouTube: youtube.com/c/ActiveInference
X: https://x.com/InferenceActive
Active Inference Livestreams: https://video.activeinference.institute/
------
CHAPTERS
0:00 Welcome and introductions
3:00 Integrative and cognitive complexity
4:30 Central prediction: open, pluralistic modeling
6:00 Measuring integrative complexity
7:30 Differentiation across domains
9:00 The scoring scale
12:00 Hierarchical integration
13:30 Cognitive structure and dialectical tension
15:00 Scoring participants
16:30 Performative complexity in elite decision-making
18:00 Automated scoring (AutoIC)
19:30 Precision allocation
21:00 Two time scales and temporal depth
22:30 Integrative complexity as abstract depth
24:00 Differentiation as competing hypotheses
25:30 Integration as structure learning
28:30 Language as an expression of the model
31:30 Responding to new problems
33:00 Model reduction and consolidation
34:30 Multi-agent settings and performativity
36:00 Elaborative complexity: a converged model
40:30 Dialectical complexity: holding alternatives
45:00 Understanding contradictory views
46:30 Elaborative vs. dialectical profiles
49:30 Task designs
51:00 Counterfactual deliberation
52:30 Complexity as state-dependent precision
54:00 Extremism and converged models
57:00 Encouraging both forms of complexity
1:01:30 Limitations and empirical tests
1:04:30 Discussion
1:09:00 Discussion: democracies over long time scales
1:13:30 Discussion: agents that model each other
1:16:30 Discussion: structure learning and precision
1:21:00 Discussion: automatic scoring and model fitting
1:24:00 Discussion: gradient-free political landscapes
1:27:00 Discussion: Dirichlet counts and coarse-grained hypotheses
1:33:00 Discussion: context and changing worlds
1:36:00 Discussion: voting and multi-agent games
1:40:30 Discussion: context-sensitive preferences
1:43:30 Discussion: neuromodulators










