On the causality between affective impact and coordinated human-robot reactions

Kasper St{\o}y; Morten Roed Frederiksen

REVIEW 3 major objections 4 minor 1 references

On the causality between affective impact and coordinated human-robot reactions

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that a small robot's reaction to a shared event alters perceived affective impact, with a reaction delay near 200 ms maximizing that impact and a delay near 100 ms maximizing the observer's felt influence on the robot.

desk verdict Plausible contingency result is undermined by a confounded delay-sweep in Study 2; the millisecond claims need stronger data. read the letter →

arxiv 2508.04834 v1 pith:VEVRT4PQ submitted 2025-08-06 cs.RO

Morten Roed Frederiksen , Kasper St{o}y This is my paper · ORCID

classification cs.RO

keywords human-robotinteractionaffectiveimpactreactioncontingencytimingsmallnon-humanoidrobotperceivedinfluencesocialrobotics

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a robot's emotional effect on people is causally shaped by how the robot reacts to a shared event, and by how fast it reacts. Two experiments with small non-humanoid robots provide the evidence: a test/control comparison with 84 observers finds a statistically significant difference ($p<.05$) between reactions tied to a shared event and random reactions, and a delay study with 110 participants locates the strongest perceived impact near 200 ms, with observers feeling most influential on the robot near 100 ms. The point of establishing this is that reaction contingency and timing become concrete, tunable design choices in human-robot interaction rather than incidental implementation details.

What carries the argument

Reaction contingency and reaction latency are the manipulated variables. The first test isolates the reactive element by comparing a robot reaction linked to an event shared with the observer against a reaction chosen at random; the second test times that contingency by raising the delay by one step for each successive group of ten participants. Together they make the robot's response policy—what event triggers a reaction and how quickly—the causal lever tested against self-reported affective impact.

What would settle it

Randomize 110 or more participants across delay conditions, or use within-subject counterbalancing, with at least ten ratings per delay and per-delay confidence intervals. If the mean perceived-impact ratings no longer show a distinct maximum near 200 ms, or perceived influence near 100 ms, the paper's millisecond conclusions are not supported. The minimal observation to look for is non-overlapping confidence intervals between neighboring delay levels.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a robot's affective impact is partly caused by whether and when its reaction is coordinated with a human observer. In the first study, 84 observers rated a small non-humanoid robot; the robot's reaction to a shared event produced a statistically significant change ($p<.05$) in perceived affective impact compared with a random reaction. In the second study, 110 participants experienced reaction delays that increased with every ten participants; the ratings trace a curve with two optima: roughly 200 ms for how much impact the robot has on the observer, and roughly 100 ms for how much impact the observer feels they have on the robot. The authors conclude fr

Load-bearing premise

The second study's millisecond peaks rest on the assumption that ten participants per delay level, assigned by enrollment order, give stable mean ratings, and that the rating scale measures affective impact as an interval quantity.

Editorial extensions

If this is right

Robot designers can treat reaction contingency as a causal control: reactions tied to shared events should raise perceived affective impact relative to arbitrary reactions.
For small non-humanoid robots, a reaction delay near 200 ms is the paper's recommended setting when the goal is the robot's impact on the observer.
When the goal is to make the observer feel influential over the robot, the paper's recommended setting is nearer 100 ms.
Near-human reaction latencies, rather than immediate or long delays, are the appropriate design target for shared physical interaction with this robot class.
Perceived affective impact is measurable enough to respond to a single manipulated parameter, reaction delay.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural next experiment is a within-subject replication with randomized delay order; it would quantify how much of the 100/200 ms effect is latency rather than order or drift.
If the timing optima hold, similar timing curves could be measured for voice, light, or motion responses, and for outcomes like trust and perceived competence.
The 100–200 ms window resembles human sensorimotor reaction times, suggesting the peak may track perceived humanness rather than an absolute delay; a more humanoid robot body might shift the optimum.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: the abstract's central claim—contingent, tightly timed reactions increase perceived affective impact—is plausible and consistent with prior HRI work. But the second study's design, as described, can't support the headline optima of ~200ms and ~100ms. Delay level is perfectly confounded with enrollment order (ten participants per level), with no error bars, effect sizes, or correction for multiple comparisons. That's a load-bearing weakness, not a quibble.

What's genuinely useful: the paper targets a concrete design question—what reaction latency should a small non-humanoid robot use—and gives a practical answer. That's a real contribution to the subfield, even if the effect itself (contingency matters) is already established. The first study, n=84 split into two groups, is a reasonable simple design, assuming the full methods actually randomized assignment and reported effect sizes.

The soft spots: I could only assess the abstract because the supplied full text is corrupted (it even contains a header from a superconductivity paper). So I can't verify the methods. That cuts both ways: the authors might have randomized within delay blocks or analyzed cohort drift, and the abstract just doesn't say. If they did, the timing conclusions could be fine. But as described, the sequential design means any gradual change in participants, experimenter behavior, or robot performance across the 110 sessions could produce exactly the kind of non-monotonic pattern they report. Ten people per point is also thin for a stable mean. The self-report interval-scale assumption is a minor worry relative to that.

Bottom line: this paper deserves a serious referee, but mainly to force the authors to show the per-delay data with confidence intervals and to prove the order effect doesn't explain the peaks. If Study 2 collapses under that test, the contribution reduces to Study 1, which is still worth publishing as a small replication plus a guarded design heuristic.

My recommendation: send it to peer review. It's not a desk reject—there's a real research question and the first experiment seems sound. But the second study needs heavy scrutiny, and the authors should be asked to reanalyze or rerun with proper randomization.

Referee Report

3 major / 4 minor

Summary. The paper reports two human-robot interaction experiments. Study 1 compares perceived affective impact when a small non-humanoid robot reacts to an event shared with a human observer versus reacting at random, with n=84 participants split into test and control groups; the abstract reports a statistically significant (p<.05) difference. Study 2 exposes 110 participants to increasingly longer reaction delays, with delay level changed every ten participants, and concludes that a delay of approximately 200 ms maximizes affective impact while approximately 100 ms best makes observers feel they influenced the robot. The paper interprets these results as evidence that reaction contingency and its timing causally shape perceived affective impact in small-robot interaction.

Significance. If the reported effects are real, the contingency manipulation and the proposed millisecond design guidelines would be practically useful for robot behavior design and would extend prior work on affective human-robot interaction. The first study's between-group comparison of shared versus random reactions is a simple, falsifiable design, and the paper clearly targets a measurable outcome. However, the second study's design and reporting are not currently adequate to support the millisecond-level claims, and the first study's abstract-level reporting omits effect sizes and test details. The contribution is potentially meaningful but needs substantially stronger empirical support.

major comments (3)

[Study 2 / Abstract] Delay level is perfectly confounded with enrollment order: the abstract states delays were increased 'with every ten participants,' giving roughly n=10 per delay level. The paper reports no randomization, counterbalancing, error bars, effect sizes, or multiple-comparison correction. Observed peaks at ~200 ms and ~100 ms are therefore indistinguishable from cohort drift or sampling noise. Because these peaks drive the paper's central timing conclusions, this is a load-bearing design problem.
[Study 1 / Abstract] The central claim that shared reactions increase perceived affective impact rests on a single p<.05 value with no test statistic, effect size, confidence interval, or description of the self-report measure's scale and validity. The practical significance of the difference cannot be assessed, and the claimed 'statistically significant change' is not enough to establish the strong causal wording in the abstract.
[Conclusions] The conclusions assert that 'a delay time around 200ms may render the biggest impact' and 'a slightly shorter reaction time around 100ms' is most effective for felt influence. These claims are stated as design recommendations despite the Study 2 confound described above. At minimum, the claims need to be reframed as exploratory, or replaced by a reanalysis with session-order controls and adequate error quantification.

minor comments (4)

[Abstract / Study 2] The abstract says 'near-human reaction times are most appropriate' but does not define 'near-human' or report the delay grid values. Please list the exact delays tested.
[Figures/Tables] The provided text contains garbled or unreadable figures and tables. If this reflects the submitted PDF, the paper needs a clean version with readable axis labels, error bars, and sample sizes in captions.
[General] The manuscript text contains an unrelated arXiv identifier and repeated blocks; these should be removed or corrected if present in the submission.
[Study 1 procedure] The 'reacting at random' control condition needs more detail: was the timing/selection truly random, and what distribution was used? This matters for interpreting the contrast with the shared-reaction condition.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; the paper's causal inference risks are experimental confounds, not circularity.

full rationale

This paper is an empirical study with two user experiments, not a derivation chain. The central claims are statistical comparisons of self-reported ratings across manipulation conditions: test versus control for shared reactions, and a sequential delay study for timing effects. There is no equation-level derivation in which an output is defined in terms of an input, no parameter fitted to a subset of data and then reported as a prediction, and no load-bearing citation to the authors' prior uniqueness theorems or ansatz-carrying prior work. The abstract reports a between-group test (n=84) and a sequential delay study (n=110) using increasingly longer delays every ten participants, but the sequential enrollment confound with delay is a threat to causal inference about the 100 ms and 200 ms peaks, not a form of derivational circularity. Likewise, the outcome is self-reported by observers who experience the manipulation, which is a construct-validity concern rather than a case of the conclusion being equivalent to its input by construction. No self-citation is visible in the provided text. Accordingly, no circular step can be quoted or exhibited, and the honest finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

Abstract-only ledger reconstructed from the abstract because the supplied full text is corrupted. The paper introduces no new physical or conceptual entities; its claims rest on the measurement validity of affective self-report, the assumption that delay is the only varied factor in study 2, and conventional significance testing. The reported 100/200 ms optima are data-fitted peak locations selected from a coarse delay grid, not theory-derived values.

free parameters (2)

Reported optimal reaction delays = ~200 ms (max affective impact), ~100 ms (max perceived influence on robot)
Empirical peak locations selected from the study-2 delay sweep with about ten participants per delay level. The abstract presents them as conclusions with no error bars or fitted model, so they function as data-fitted values for the paper's headline claims.
Delay grid levels in study 2 = unspecified in abstract; incremented every ten participants from short to long
Experimenter-chosen independent variable; its coarseness and sequential ordering limit the resolution and clean comparison of the reported optima.

assumptions (3)

domain assumption Self-reported ratings of affective impact and felt impact on the robot are valid interval-scale measures of the intended psychological constructs.
The central outcome is a rating scale; the abstract does not describe validation, item wording, or scale properties. This is the measurement premise on which the p<0.05 result rests.
domain assumption In study 2, programmed reaction delay is the only meaningfully varied factor between the ten-person cohorts, with otherwise identical robot behavior.
The abstract's delay sweep treats per-cohort differences in ratings as caused by delay; any confound (enrollment-order effects, participant composition drift over time, experimenter or hardware drift) would corrupt the 100/200 ms conclusions.
domain assumption Standard frequentist null-hypothesis testing at alpha = 0.05 is appropriate, and no multiple-comparison correction is needed across the delay levels and outcome measures.
The abstract reports p<0.05 without mentioning corrections; this is customary but load-bearing for the statistical claims, especially given the multi-level delay sweep.

how reviews work

0 comments

Cite this review

Pith. "Pith review of On the causality between affective impact and coordinated human-robot reactions." pith.science (2026). https://pith.science/paper/VEVRT4PQ

@misc{pith2026250804834,
  author       = {Pith},
  title        = {Pith review of: On the causality between affective impact and coordinated human-robot reactions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VEVRT4PQ}},
  note         = {Machine review of arXiv:2508.04834}
}

abstract

In an effort to improve how robots function in social contexts, this paper investigates if a robot that actively shares a reaction to an event with a human alters how the human perceives the robot's affective impact. To verify this, we created two different test setups. One to highlight and isolate the reaction element of affective robot expressions, and one to investigate the effects of applying specific timing delays to a robot reacting to a physical encounter with a human. The first test was conducted with two different groups (n=84) of human observers, a test group and a control group both interacting with the robot. The second test was performed with 110 participants using increasingly longer reaction delays for the robot with every ten participants. The results show a statistically significant change (p$<$.05) in perceived affective impact for the robots when they react to an event shared with a human observer rather than reacting at random. The result also shows for shared physical interaction, the near-human reaction times from the robot are most appropriate for the scenario. The paper concludes that a delay time around 200ms may render the biggest impact on human observers for small-sized non-humanoid robots. It further concludes that a slightly shorter reaction time around 100ms is most effective when the goal is to make the human observers feel they made the biggest impact on the robot.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

1 extracted references · 1 canonical work pages

[1]

� �� ...

work page Pith review arXiv 2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

� �� ...

work page Pith review arXiv 2025