Better Safe than Sorry? - An Active Inference Approach to Biased Social Inference in Depression DOI Open Access
Lukas Kirchner, Anna-Lena Eckert, Max Berg

et al.

Published: Aug. 9, 2022

Individuals with depressive disorders reveal marked distortions in their social perception and behavior. Self-reinforcing vicious cycles of avoidance increasing anxiety can negatively influence the disease’s course. Clinical psychology has offered many explanations as to why these tend persist depression, even when patient’s context changes for better. Active inference, a general computational theory perception, planning, behavior, potential improve psychological models depression. Its flexible mathematical formalization offers new avenues towards understanding underlying mechanisms by modelling implicit, inferential processes. We argue that maintenance symptoms is primarily due how (and what model world) people depression infer nature contexts through action (e.g., decision making). In line recent work on processes, we propose conceptualize inference partially observable Markov process (POMDP). This allows us formalize different “phenotypes” processing behavior For example, overly precise, negative prior beliefs about hidden state may trigger more pessimistic making whereas very imprecise should result insecure behaviors. Finally, outline research agenda suggest relevant applications diagnostics treatment selection.

Language: Английский

Geometric methods for sampling, optimization, inference, and adaptive agents DOI
Alessandro Barp, Lancelot Da Costa, Guilherme França

et al.

Handbook of statistics, Journal Year: 2022, Volume and Issue: unknown, P. 21 - 78

Published: Jan. 1, 2022

Language: Английский

Citations

8
An Overview of the Free Energy Principle and Related Research DOI
Zhengquan Zhang, Feng Xu

Neural Computation, Journal Year: 2024, Volume and Issue: 36(5), P. 963 - 1021

Published: March 8, 2024

Abstract The free energy principle and its corollary, the active inference framework, serve as theoretical foundations in domain of neuroscience, explaining genesis intelligent behavior. This states that processes perception, learning, decision making—within an agent—are all driven by objective “minimizing energy,” evincing following behaviors: learning employing a generative model environment to interpret observations, thereby achieving selecting actions maintain stable preferred state minimize uncertainty about environment, making. fundamental can be used explain how brain perceptual information, learns selects actions. Two pivotal tenets are agent employs for perception planning interaction with world (and other agents) enhances performance augments perception. With evolution control theory deep tools, agents based on FEP have been instantiated various ways across different domains, guiding design multitude models decision-making algorithms. letter first introduces basic concepts FEP, followed historical development connections theories intelligence, then delves into specific application making, encompassing both low-dimensional simple situations high-dimensional complex situations. It compares model-based reinforcement show provides better function. We illustrate this using numerical studies Dreamer3 adding expected information gain standard In complementary fashion, existing algorithms also help implement FEP-based agents. Finally, we discuss capabilities need possess environments aid acquiring these capabilities.

Language: Английский

Citations

1
Hesitation, orientation, and flow: A taxonomy for deep temporal translation architectures DOI Creative Commons
Michaël Carl, Yuxiang Wei, Sheng Lu

et al.

Ampersand, Journal Year: 2024, Volume and Issue: 12, P. 100164 - 100164

Published: Feb. 2, 2024

Numerous models have been proposed to describe and predict how human translations evolve in time. Some of these suggest hierarchically embedded processes fast slow processing that unfold on different timelines. However, the assumed mental often conceptualized without a clear description they can be assessed, measured or retrieved behavioral data. Other approaches fragmenting data into various kinds units, but status units with respect their cognitive reality is not always very clear. In this paper, we propose novel annotation taxonomy for data, assuming three broad states translators experience during translation production: A state orientation (O) reflects epistemic foraging which translator reads scans piece source text (ST) searches information. flow (F), engages fluent production characterized by focus involvement process. hesitation (H) described terms uncertainty doubt results patterns re-reading, modification disfluent production. This HOF aims at describing elicit experiential qualities process, associated typical recorded process (TPD, i.e., logged keystrokes gaze data). We manual small set TPD develop method ensures high inter-rater agreement (kappa 0.88). show our cluster higher-level strategies (so-called policies). discuss policies are optimized trigger off lower-level processes. compare other deep-temporal architecture assumes hierarchy interact ways.

Language: Английский

Citations

1
On Predictive Planning and Counterfactual Learning in Active Inference DOI Creative Commons
Aswin Paul, Takuya Isomura, Adeel Razi

et al.

Entropy, Journal Year: 2024, Volume and Issue: 26(6), P. 484 - 484

Published: May 31, 2024

Given the rapid advancement of artificial intelligence, understanding foundations intelligent behaviour is increasingly important. Active inference, regarded as a general theory behaviour, offers principled approach to probing basis sophistication in planning and decision-making. This paper examines two decision-making schemes active inference based on “planning” “learning from experience”. Furthermore, we also introduce mixed model that navigates data complexity trade-off between these strategies, leveraging strengths both facilitate balanced We evaluate our proposed challenging grid-world scenario requires adaptability agent. Additionally, provides opportunity analyse evolution various parameters, offering valuable insights contributing an explainable framework for

Language: Английский

Citations

1
An Active Inference Agent for Modeling Human Translation Processes DOI Creative Commons
Michaël Carl

Entropy, Journal Year: 2024, Volume and Issue: 26(8), P. 616 - 616

Published: July 23, 2024

This paper develops an outline for a hierarchically embedded architecture of artificial agent that models human translation processes based on principles active inference (AIF) and predictive processing (PP). AIF PP posit the mind constructs model environment which guides behavior by continually generating integrating predictions sensory input. The proposed consists three strata: sensorimotor layer, cognitive phenomenal layer. Each layer network states transitions interact different time scales. Following framework, are conditioned observations may originate from and/or while between actions implement plans to optimize goal-oriented behavior. aims at simulating variation in translational under various conditions facilitate investigating underlying mental mechanisms. It provides novel framework testing new hypotheses translating mind.

Language: Английский

Citations

1
Associative Learning and Active Inference DOI

Petr Anokhin,

Artyom Sorokin,

Mikhail Burtsev

et al.

Neural Computation, Journal Year: 2024, Volume and Issue: 36(12), P. 2602 - 2635

Published: Sept. 23, 2024

Abstract Associative learning is a behavioral phenomenon in which individuals develop connections between stimuli or events based on their co-occurrence. Initially studied by Pavlov his conditioning experiments, the fundamental principles of have been expanded through discovery wide range phenomena. Computational models developed concept minimizing reward prediction errors. The Rescorla-Wagner model, particular, well-known model that has greatly influenced field reinforcement learning. However, simplicity these restricts ability to fully explain diverse phenomena associated with In this study, we adopt free energy principle, suggests living systems strive minimize surprise uncertainty under internal world. We consider process as minimization and investigate its relationship focusing informational aspects learning, different types surprise, errors beliefs values. Furthermore, explore how such blocking, overshadowing, latent inhibition can be modeled within active inference framework. accomplish using novelty attention, share similar ideas proposed seemingly contradictory Mackintosh Pearce-Hall models. Thus, demonstrate theoretical framework derived from first principles, integrate associative empirical experiments serve for better understanding computational processes behind brain.

Language: Английский

Citations

1
A hierarchical active inference model of spatial alternation tasks and the hippocampal-prefrontal circuit DOI Creative Commons
Toon Van de Maele, Bart Dhoedt, Tim Verbelen

et al.

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Nov. 15, 2024

Cognitive problem-solving benefits from cognitive maps aiding navigation and planning. Physical space involves hippocampal (HC) allocentric codes, while abstract task engages medial prefrontal cortex (mPFC) task-specific codes. Previous studies show that challenging tasks, like spatial alternation, require integrating these two types of maps. The disruption the HC-mPFC circuit impairs performance. We propose a hierarchical active inference model clarifying how this solves interaction tasks by bridging physical task-space Simulations demonstrate model's dual layers develop effective for space. alternation through reciprocal interactions between layers. Disrupting its communication decision-making, which is consistent with empirical evidence. Additionally, adapts to switching multiple rules, providing mechanistic explanation supports effects disruption. How interact when executing not fully understood. This paper models hippocampal-prefrontal circuits memory-guided taskspace.

Language: Английский

Citations

1
Learning in Hybrid Active Inference Models DOI

Poppy Collis,

Ryan Singh,

Paul Kinghorn

et al.

Communications in computer and information science, Journal Year: 2024, Volume and Issue: unknown, P. 49 - 71

Published: Dec. 30, 2024

Language: Английский

Citations

1
Learning Generative Models for Active Inference Using Tensor Networks DOI
Samuel T. Wauthier, Bram Vanhecke, Tim Verbelen

et al.

Communications in computer and information science, Journal Year: 2023, Volume and Issue: unknown, P. 285 - 297

Published: Jan. 1, 2023

Language: Английский

Citations

3
Simulating Active Inference of Interpersonal Context Within and Across Mental Disorders DOI Open Access
Anna-Lena Eckert,

Janik Pawlowski,

Winfried Rief

et al.

Published: Oct. 18, 2023

Background. Interpersonal problems are common in mental disorders like depression and social anxiety disorder. However, there is a lack of formal models to explain idiosyncrasies patients' interpersonal functioning. Following modern computational accounts perception action, experience behavior results from implicit inference about hidden environmental properties.Methods. We simulate decision-making “trust game” using POMDP generative model active inference. Simulated agents decide either keep or invest an initial budget trustee. If the invests, trustee can cooperate with exploit agent’s trust. Agents perform context (cooperative vs. hostile) action selection. By introducing systematic biases model, we several subtypes anxiety. then collected data N=25 patients diagnosed depressive fit data. Results. biased expectations preferences showed idiosyncratic They more readily infer be hostile avoid investing Biases also affected total earned rewards, socially anxious depressed receiving higher average rewards contexts compared "healthy" (i.e., accurate-to-optimistic model), but “healthy” top optimistic showing superior performance volatile environments. Fitting resulted individual parameters for low symptom scores those high scores, group differences particularly transition dynamics B, outcome C prior beliefs D. Discussion. Our simulations formalize complex phenomena within Active Inference. formalized investigated respect their functional role. The has potential applications psychopathological research, personalized diagnostics, individualized treatment planning. Further empirical work necessary validate basis clinical usefulness decision-making.

Language: Английский

Citations

3