The DOI matches its Crossref record; 68 of the 71 references it cites were found in a registry or an index; 10 of the 39 sources its claims cite were checked for support.
Understanding Agent Complexity Through Affordances
Open accessAbstract
The rapid development of autonomous systems, some of which emulate animal abilities, raises a question about how to understand and compare their complexity. Taking an “agent” to be any system that can be well-explained by attributing goals and intentional states to it, in what sense should we understand some agents as more “complex” than others? There is a sizable literature about complexity in biology, notably in comparative psychology, where an animal’s complexity informs predictions about its abilities and guides experimental design. In engineering and artificial intelligence, complexity is often assumed to be a sign of progress – we want to build more complex, capable machines. While the goals of researchers in these fields are different, they plausibly overlap in common conceptions of what a “more complex agent” entails. Existing work on “environmental complexity” and its relationship to “complex cognition” inquires into the origins of complexity, but not what it means to be a complex agent. Recognizing that complex behavior comes not just from complex cognition but from the interplay of body, environment, and cognition, we articulate an account of agent complexity in terms of Gibson’s (The Ecological Approach to Visual Perception (Classic), Houghton Mifflin, Boston, 1979) notion of affordances. We distinguish between cognitive and behavioral dimensions of affordances, and offer a framework for comparing agents based on the range of affordances they can exploit. Our approach ecologically grounds the sense in which some agents are more complex than others, while also clarifying why the answer to “which agent is more complex?” is sometimes indeterminate. It enables systematic comparison across diverse agent types, as we demonstrate through a detailed analysis of the nematode C. elegans and the hexapod robot RHex. We hope our account can advance discourse around the complexity of animals and guide the development of complex, agential machines.
Overview
CoChat Summary
TL;DR
The paper proposes an affordance-based account of agent complexity, where agents are compared by the range of behavioral and cognitive affordances they can exploit rather than by intelligence, human-likeness, or centralized computation. It illustrates the framework by comparing the nematode *C. elegans* and the hexapod robot RHex, arguing that *C. elegans* is more complex both behaviorally and cognitively.
Key findings
- Agent complexity is analyzed as a system’s current modes of activity, distinguished from intelligence, and grounded in Gibsonian affordances: an affordance is a possibility for an agent’s activity in a given environment.
- The framework separates behavioral and cognitive affordances and offers sufficient conditions for comparing capacities: capacity A is more complex than B if A’s basic affordances are a superset of B’s, or if they can be pairwise compared so that no B-affordance is more complex than its paired A-affordance and at least one A-affordance is more complex.
- In the *C. elegans* vs. RHex comparison, *C. elegans* has a broader repertoire of locomotor modes, more sensory modalities, and multiple forms of learning; base RHex has no composite cognitive affordances, while augmented RHex gains locomotor and navigation affordances.
- Even augmented RHex is less complex by the paper’s enumeration: *C. elegans* possesses “nearly 50% more distinct affordances” than augmented RHex, with many worm affordances addressing dimensions outside RHex’s domain, such as chemicals and reproduction; *C. elegans* has a small nervous system of 302 neurons.
Methods
- Conceptual and philosophical analysis: the paper develops an affordance-based framework from ecological psychology, the Dennettian intentional stance, and existing complexity debates including the Environmental Complexity Thesis, the Zero Force Evolutionary Law, and Mikhalevich et al. on behavioral flexibility.
- Illustrative comparative case study: the authors enumerate and organize affordances for *C. elegans* and RHex into basic/composite and behavioral/cognitive categories, using the base RHex model and an augmented version with added sensors and code.
- Comparative analysis: the framework is applied to the *C. elegans*–RHex case and contrasted with Cochet and Byrne’s three-dimensional account of complexity (motor precision, coordination, planning).
Limitations
- Stated: the *C. elegans* affordance taxonomy involves interpretive choices and is not authoritative; a specialist might categorize its affordances differently, and the listing is provisional because new behaviors continue to be discovered.
- Stated: the paper does not defend an ultimate answer to the behavioral–cognitive distinction and relies on the reader’s intuitive view of that divide; it also leaves open exactly which affordances are basic.
- Stated and evident: the account is silent about some comparison cases where the capacity with fewer component affordances has at least one more complex affordance, and intermediate cases may lack a determinate answer; the authors also note that their broad notion of agency might obscure significant characteristics of animal agency.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
6 Valid · 5 Not valid · 16 Not checked · 94 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Octopuses are an instructive example, having diverged from our genetic line even before the development of vertebral columns, yet demonstrating far more in the way of flexible, innovative problem solving than any frog or lizard (Schnell et al., 2020).”
What the sources say
- Partly supportedcites Schnell et al., 2020 sourceExperimental — accuracy not yet measured
Section III (page unavailable) describes cephalopods, including octopuses, as behaviourally flexible and able to modify their behaviour innovatively. The paper also says cephalopods diverged from the vertebrate lineage over 550 million years ago (Patterns of connectivity; page unavailable). It does not establish that octopuses show more flexible, innovative problem solving than any frog or lizard. (LLM, conf=0.88)
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“a low center of mass, high friction feet, and more than two legs are morphological adaptations that combine well with prediction-based locomotion strategies like MPC by reducing the difficulty of the prediction problem”
What the sources say
- Could not verifycites Johnson et al., 2012 sourceExperimental — accuracy not yet measured
Our copy of Johnson, Haynes & Koditschek (2012) never addresses prediction-based locomotion or MPC: the Abstract and Section I present a deliberately model-free, quasi-static 'reactive stand' that minimizes internal and external torques 'absent any terrain information' and 'without recourse to further modeling'. The closest morphology-related statements are incidental assumptions — terrain with 'sufficient friction and reasonable surface normals' (Sec. III-A, Problem Setup) and the generalization to X-RHex's six legs (Sec. III-D, Implementation on a Spatial Hexapod, which notes extra legs add redundancy but also a higher chance of legs lifting off) — neither tied to easing a prediction problem. Searches for 'prediction/model predictive control/MPC', 'low center of mass, high friction feet, more than two legs', 'static stability morphology', and 'model-free vs model-based' returned no passage supporting or contradicting the attributed claim, so the claim's point lies outside this paper. (LLM, conf=0.70)
- Not matchedcites Li et al., 2009
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Without being able to rely as others do on the vestibular system, when he runs, he must coordinate information from multiple other sensory systems.”
What the sources say
- Partly supportedcites Sanlaville and Verloes, 2007 sourceExperimental — accuracy not yet measured
The Auditory and vestibular anomalies section describes absent semicircular canals and impaired vestibular responses in CHARGE syndrome. The Motor milestones section links delayed movement to vestibular and visual impairments. Neither section says that affected individuals coordinate information from multiple other sensory systems when running; the paper also describes inner-ear malformations, not necessarily the absence of the entire inner ear. (LLM, conf=0.85)
plainstates a causeSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“our enumeration shows that C. elegans possesses nearly 50% more distinct affordances than even the augmented RHex”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“In short, we propose that reasonable comparative judgments about complexity can be based on the range of affordances available to an agent in its environment.”
What the sources say
- Not matchedcites Gibson, 1966, 1979
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
39 more claims
Create a free account to see all 121 claims and to verify the 1 reference still open.
- Probably not a substantial claim. 59
- It may state two separate claims as one. 28
- The quote only partly states this claim. 1
Citations
Cites
No free lunch theorems for optimization
Wolpert, D. H., & Macready, W. G. (1997) · IEEE Transactions on Evolutionary Computation · doi:10.1109/4235.585893 · Cited by 14,237 · cited 1×
“This is partly because “intelligence” is a shifty concept, whose meaning depends on what problems one cares about (Wolpert and Macready, 1997; Sternberg and Grigo-renko, 2004; Curry, 2021 ; Baker, 2025).”
Reducibility among Combinatorial Problems
Karp, R. M. (1972) · Complexity of Computer Computations · doi:10.1007/978-1-4684-2001-2_9 · Cited by 11,511
An introduction to cybernetics
Ashby, W. R. (1956) · doi:10.5962/bhl.title.5851 · Cited by 7,578 · cited 1×
“This emphasis on agent-environment coupling echoes earlier cybernetic approaches (Ashby, 1956; Wiener, 1948 ), though Gibson’s affordances offer a more action-oriented framework for understanding these interactions.”
The complexity of theorem-proving procedures
Cook, S. A. (1971) · Proceedings of the third annual ACM symposium on Theory of computing - STOC '71 · doi:10.1145/800157.805047 · Cited by 6,385 · cited 1×
“That said, computational complexity theory can be used to classify types of problems, which makes it somewhat more relevant to our concern (Cook, 1971; Karp, 1972; Sipser, 2012).”
What we can do and what we cannot do with fMRI
Logothetis, N. K. (2008) · Nature · doi:10.1038/nature06976 · Cited by 3,612 · cited 1×
“This distinction is prominent, for example, in neuroscientific research on apes and humans that aims to measure covert activ - ity relating to perception or attention (Logothetis, 2008; Luck, 2005).”
Building machines that learn and think like people
Lake, B. M., Ullman, T. D., Tenenbaum, J. B., & Gershman, S. J. (2017) · Behavioral and Brain Sciences · doi:10.1017/s0140525x16001837 · Cited by 2,550
An Introduction to Kolmogorov Complexity and Its Applications
Li, M., & Vitányi, P. (2008) · Texts in Computer Science · doi:10.1007/978-0-387-49820-1 · Cited by 1,424 · cited 1×
“Among the formal approaches to complexity that have been developed in computer science, Kolmogorov complexity and the related notion of logical depth have been influential (Kolmogorov, 1965; Li & Vitányi, 2008 ; Bennett, 1988).”
RHex: A Simple and Highly Mobile Hexapod Robot
Saranli, U., Buehler, M., & Koditschek, D. E. (2001) · The International Journal of Robotics Research · doi:10.1177/02783640122067570 · Cited by 1,322 · cited 1×
“An example of using morphological adaptation strategies can be seen in RHex, short for Robot Hexapod (Saranli et al., 2001; Galloway et al., 2010).”
Closed-loop brain training: the science of neurofeedback
Sitaram, R., Ros, T., Stoeckel, L., Haller, S., Scharnowski, F., Lewis-Peacock, J., Weiskopf, N., Blefari, M. L., Rana, M., Oblak, E., et al. (2017) · Nature Reviews Neuroscience · doi:10.1038/nrn.2016.164 · Cited by 1,306 · cited 1×
“Such a person engages in what we would normally call cognitive activity (say, visualizing or remembering something), whose goal is to move an object or win a game (Sitaram et al., 2017).”
A Rich Landscape of Affordances
Rietveld, E., & Kiverstein, J. (2014) · Ecological Psychology · doi:10.1080/10407413.2014.958035 · Cited by 1,160 · cited 1×
“Our view captures this in the way that motivational capacities directly determine affordances: an agent with self-restraint is afforded delayed rewards that remain inaccessible to otherwise similar agents lack - ing this capacity (Gallagher, 2007; Rietveld and Kiverstein, 2014).”
Research works printed but not linked: 9
- Baker, B. (2025). Rethinking artificial general intelligence. In D. Curry & L. Daoust (Eds.), Introducing Philosophy of Mind. London and New York: Today. Routledge. Forthcoming.
- Browning, J.: Intentionality all-stars redux: Do language models know what they are talking about? In: Cappelen, H., Sterken, R. (eds.) Communicating with AI: Philosophical Perspectives. Oxford University Press, ??? ( forthcoming)
- Dennett, D. C. (1987). The Intentional Stance. MIT Press.
- Gibson, J. J. (1966). The Senses Considered as Perceptual Systems. Houghton Mifflin.
- Godfrey-Smith, P. (2001). Environmental complexity and the evolution of cognition. In R. J. Sternberg & J. C. Kaufman (Eds.), The Evolution of Intelligence (pp. 233–249). Mahwah, NJ: Lawrence Erlbaum.
- Kolmogorov, A.N (1965) Three approaches to the quantitative definition of information. Problemy Peredachi Informatsii 1(1), 3–11,. In Russian. English translation in Problems of Information Transmission,1(1), 1–7.
- Kuindersma, S., Valenzuela, A., Ferrari, F., Mastalli, C., & Ponton, B. (2024). Advances in humanoid model predictive control. Annual Review of Control, Robotics, and Autonomous Systems, 7, 123–148. https://doi.org/10.1146/annurev-control-072523-111032 · doi:10.1146/annurev-control-072523-111032
- Luck, S. J. (2005). An Introduction to the Event-Related Potential Technique. MIT Press.
- Sipser, M. (2012). Introduction to the Theory of Computation (3rd ed.). Cengage Learning.
Metadata
- Authors
- Baker, B., & Roberts, S.
- Affiliations
- Colby College (US)
- Wesleyan University (US)
- Corresponding author
- Baker, Ben
- Venue
- Minds and Machines
- Published
- 2025-11-13
- Volume
- 35
- Issue
- 4
- ISSN
- 0924-6495 (ISSN-L)
- 1572-8641
- OpenAlex
- W4416162083
- Semantic Scholar
- 1754bd2e22026cd3d944f5adda716bf90caee4db
- Cited by
- 0
- References
- 61
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by-nc-nd
- creativecommons.org/licenses/by-nc-nd/4.0 (tdm)
- creativecommons.org/licenses/by-nc-nd/4.0 (vor)
- Locations
- Minds and Machines (publishedVersion, cc-by-nc-nd)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Embodied and Extended Cognition
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Cybernetics and Technology in Society
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Philosophy of Biology and Scientific Explanation
- Keywords
- affordances, ecological psychology, complex cognition, environmental complexity, comparative cognition, animal cognition, Caenorhabditis elegans, autonomous systems, intentional states
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Understanding Agent Complexity Through Affordances. https://app.cochat.ai/papers/10.1007/s11023-025-09748-7
- The paper
- Baker, B., & Roberts, S. (2025). Understanding Agent Complexity Through Affordances. Minds and Machines. https://doi.org/10.1007/s11023-025-09748-7
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.