Beyond Theory of Mind in Robotics

Malte F. Jung

arxiv: 2604.09612 · v1 · submitted 2026-03-12 · 💻 cs.AI · cs.HC

Beyond Theory of Mind in Robotics

Malte F. Jung This is my paper

Pith reviewed 2026-05-15 11:41 UTC · model grok-4.3

classification 💻 cs.AI cs.HC

keywords Theory of Mindsocial roboticscoordinationethnomethodologyconversation analysisparticipatory sense-makinghuman-robot interaction

0 comments

The pith

Social meaning in robot interactions arises from real-time coordination rather than inferred hidden mental states.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper contends that Theory of Mind approaches in robotics rest on assumptions that do not match how social interactions typically occur. These assumptions include meaning originating from inside agents, requiring passive inference to understand, and having fixed interpretations. Drawing from alternative fields, it proposes that meaning is instead produced through active, moment-to-moment coordination. This shift carries direct consequences for how robots should be designed to participate in social settings. Readers might find this relevant as it questions the basis of many current social robot systems.

Core claim

Social meaning is not decoded from behavior but produced through moment-to-moment coordination between agents. The three assumptions of ToM—that meaning travels inside-out from hidden states to observable behavior, that understanding requires detached inference rather than participation, and that the meaning of behavior is fixed and available to a passive observer—poorly capture social interaction. An alternative foundation from ethnomethodology, conversation analysis, and participatory sense-making supports designing robots around policies for sustaining coordination, active participation, and stabilizing meaning potential through response.

What carries the argument

The production of social meaning through moment-to-moment coordination between agents, which replaces internal state modeling and detached inference.

If this is right

Robot design should move from internal state modeling to policies for sustaining coordination.
Design should prioritize active participation over observer-based inference.
Behavioral meaning should be treated as potential stabilized through response rather than fixed.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

This could lead to robots that maintain longer and more natural interactions by focusing on joint actions.
It suggests connections to embodied cognition theories where understanding emerges from doing rather than thinking.
Future tests might measure coordination success rates without mental state inference in human-robot pairs.

Load-bearing premise

The three assumptions of Theory of Mind about how meaning is transmitted, understood, and fixed do not reflect the actual unfolding of most social interactions.

What would settle it

An experiment comparing interaction success and duration between robots using only coordination policies versus those relying on ToM-based inference of mental states.

read the original abstract

Theory of Mind, the capacity to explain and predict behavior by inferring hidden mental states, has become the dominant paradigm for social interaction in robotics. Yet ToM rests on three assumptions that poorly capture how most social interaction actually unfolds: that meaning travels inside-out from hidden states to observable behavior; that understanding requires detached inference rather than participation; and that the meaning of behavior is fixed and available to a passive observer. Drawing on ethnomethodology, conversation analysis, and participatory sense-making, I argue that social meaning is not decoded from behavior but produced through moment-to-moment coordination between agents. This interactional foundation has direct implications for robot design: shifting from internal state modeling toward policies for sustaining coordination, from observer-based inference toward active participation, and from fixed behavioral meaning toward meaning potential stabilized through response.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This is a clear conceptual critique that correctly flags how ToM assumptions misalign with real interaction, but it needs concrete robot examples to move from reframing to usable guidance.

read the letter

The main takeaway is that Jung argues Theory of Mind has become the default in social robotics because it treats meaning as hidden states to be inferred, when most interaction actually builds meaning through ongoing coordination. He pulls three assumptions from the literature—meaning travels inside-out, understanding needs detached inference, and behavior has fixed meaning—and shows how they steer design toward internal modeling rather than participation. That part lands cleanly and matches what the abstract and stress-test describe. The paper does well by keeping the argument short and by naming the design shifts explicitly: move from state estimation to policies that sustain coordination, from passive observation to active response, and from fixed meanings to ones stabilized in the moment. Those implications follow directly from the cited traditions in ethnomethodology and conversation analysis without adding extra claims. No circular fitting or invented entities appear. The soft spot is that everything stays at the level of principled suggestion. The three assumptions are listed but not shown to be load-bearing failures in actual robot deployments, and the alternative is described in general terms without a worked example of what a coordination policy would look like or how it would be tested against a ToM baseline. That gap makes it harder to judge whether the reframing changes engineering practice or just the surrounding talk. This paper is for researchers in human-robot interaction who already read across sociology and interaction studies. A reader looking for new algorithms or data will find little to use directly, while someone open to rethinking priorities will see a coherent counterpoint worth discussing. I would send it to peer review because the logic holds on its own terms and the field benefits from these interdisciplinary position pieces even when they remain high-level.

Referee Report

2 major / 2 minor

Summary. The paper claims that Theory of Mind (ToM) has become the dominant paradigm for social interaction in robotics yet rests on three assumptions that poorly reflect actual interaction: meaning travels inside-out from hidden states, understanding requires detached inference, and behavioral meaning is fixed for passive observers. Drawing on ethnomethodology, conversation analysis, and participatory sense-making, the manuscript argues that social meaning is instead produced through moment-to-moment coordination. It concludes that this interactional foundation implies concrete shifts in robot design: from internal state modeling to coordination-sustaining policies, from observer inference to active participation, and from fixed meanings to response-stabilized meaning potentials.

Significance. If the reframing holds, the work could redirect social robotics research toward coordination-focused and participatory architectures rather than mental-state inference. Its conceptual nature means significance would derive from stimulating new design paradigms and empirical follow-up rather than from immediate technical results or validated mechanisms.

major comments (2)

[Abstract] Abstract: The central claim that the three listed ToM assumptions 'poorly capture how most social interaction actually unfolds' is asserted without derivation, data, or mapping to specific robotics failure cases (e.g., no cited studies showing where ToM-based policies have produced coordination breakdowns in situated tasks). This assertion is load-bearing for motivating the proposed design shift.
[Design implications] Design implications paragraph: The three suggested shifts (internal modeling to coordination policies, observer inference to active participation, fixed meaning to response-stabilized potential) are stated at the level of principled suggestion without even a schematic example of a coordination policy or a contrast with an existing ToM implementation on a concrete robot task, leaving the translation from critique to design underspecified.

minor comments (2)

Add a brief table or bullet list explicitly contrasting the three ToM assumptions with the corresponding interactional alternatives to improve readability.
Ensure all references to ethnomethodology, conversation analysis, and participatory sense-making include complete bibliographic details and page numbers for key claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these constructive comments, which help clarify how to strengthen the manuscript's grounding and practical relevance. We address each point below and will revise accordingly.

read point-by-point responses

Referee: [Abstract] Abstract: The central claim that the three listed ToM assumptions 'poorly capture how most social interaction actually unfolds' is asserted without derivation, data, or mapping to specific robotics failure cases (e.g., no cited studies showing where ToM-based policies have produced coordination breakdowns in situated tasks). This assertion is load-bearing for motivating the proposed design shift.

Authors: The claim is derived from the cited bodies of work in ethnomethodology and conversation analysis, which supply extensive empirical documentation of interactional meaning production. The manuscript already references robotics contexts where detached inference approaches encounter difficulties with real-time coordination (e.g., dialogue repair and joint action). We will revise the abstract and introduction to make these mappings more explicit by adding targeted citations to studies of ToM-based robot failures in situated tasks, thereby strengthening the motivation without introducing new empirical data. revision: yes
Referee: [Design implications] Design implications paragraph: The three suggested shifts (internal modeling to coordination policies, observer inference to active participation, fixed meaning to response-stabilized potential) are stated at the level of principled suggestion without even a schematic example of a coordination policy or a contrast with an existing ToM implementation on a concrete robot task, leaving the translation from critique to design underspecified.

Authors: We agree that the implications section would be strengthened by greater concreteness. In revision we will add a brief schematic example of a coordination policy (e.g., a response-contingency rule for turn allocation that stabilizes meaning through participation rather than state inference) and contrast it directly with a standard ToM-based implementation on a simple robot dialogue task. This will illustrate the design translation while preserving the paper's conceptual focus. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper is a conceptual position piece that critiques ToM assumptions by drawing on external literatures (ethnomethodology, conversation analysis, participatory sense-making) and derives design implications from that reframing. No equations, fitted parameters, self-definitional reductions, or load-bearing self-citations appear in the provided text. The central argument rests on cited external traditions rather than reducing to its own inputs by construction. This is the expected outcome for a non-mathematical, literature-grounded argument with no internal fitting or renaming steps.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on adopting the three listed limitations of ToM and the positive claim that meaning is produced in coordination; these are treated as given from the cited fields without further justification in the abstract.

axioms (2)

domain assumption Theory of Mind rests on three assumptions (inside-out meaning, detached inference, fixed meaning) that poorly capture most social interaction
Explicitly stated as the starting point for the critique in the abstract.
domain assumption Social meaning is produced through moment-to-moment coordination between agents
Central positive claim drawn from ethnomethodology, conversation analysis, and participatory sense-making.

pith-pipeline@v0.9.0 · 5419 in / 1362 out tokens · 38056 ms · 2026-05-15T11:41:19.965625+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

[1]

Admoni, H., & Scassellati, B. (2017). Social eye gaze in human-robot interaction: a review. Journal of Human-Robot Interaction, 6(1), 25-63. Baron-Cohen, S. (1997). Mindblindness: An essay on autism and theory of mind. MIT press. Breazeal, C., Buchsbaum, D., Gray, J., Gatenby, D., & Blumberg, B. (2005). Learning from and about others: Towards using imitat...

work page 2017
[2]

Cha, E., Kim, Y ., Fong, T., & Matarić, M. J. (2018). A survey of nonverbal signal- ing methods for non-humanoid robots. Foundations and Trends® in Robotics, 6(4), 211-323. Beyond ToM Dennett, D. C. (1989). The intentional stance. MIT press. De Jaegher, H., & Di Paolo, E. (2007). Participatory sense-making: An enactive approach to social cognition. Phenom...

work page 2018
[3]

M., Harrison, A

Beyond ToM Hiatt, L. M., Harrison, A. M., & Trafton, J. G. (2011, July). Accommodating hu- man variability in human-robot teams through theory of mind. In IJCAI Proceedings-In- ternational Joint Conference on Artiﬁcial Intelligence (Vol. 22, No. 3, p. 2066). Jung, M. F . (2017). Aﬀective grounding in human-robot interaction. In Proceed- ings of the 2017 A...

work page 2011

[1] [1]

Admoni, H., & Scassellati, B. (2017). Social eye gaze in human-robot interaction: a review. Journal of Human-Robot Interaction, 6(1), 25-63. Baron-Cohen, S. (1997). Mindblindness: An essay on autism and theory of mind. MIT press. Breazeal, C., Buchsbaum, D., Gray, J., Gatenby, D., & Blumberg, B. (2005). Learning from and about others: Towards using imitat...

work page 2017

[2] [2]

Cha, E., Kim, Y ., Fong, T., & Matarić, M. J. (2018). A survey of nonverbal signal- ing methods for non-humanoid robots. Foundations and Trends® in Robotics, 6(4), 211-323. Beyond ToM Dennett, D. C. (1989). The intentional stance. MIT press. De Jaegher, H., & Di Paolo, E. (2007). Participatory sense-making: An enactive approach to social cognition. Phenom...

work page 2018

[3] [3]

M., Harrison, A

Beyond ToM Hiatt, L. M., Harrison, A. M., & Trafton, J. G. (2011, July). Accommodating hu- man variability in human-robot teams through theory of mind. In IJCAI Proceedings-In- ternational Joint Conference on Artiﬁcial Intelligence (Vol. 22, No. 3, p. 2066). Jung, M. F . (2017). Aﬀective grounding in human-robot interaction. In Proceed- ings of the 2017 A...

work page 2011