A potential mechanism for Gibsonian resonance: behavioral entrainment emerges from local homeostasis in an unsupervised reservoir network

May combine two claims.

The concept of resonance in ecological psychology remains a metaphor and lacks a mechanistic account.

“The concept of resonance has played an important role in ecological psychology, but it remains a metaphor.”

• cites Raja, 2018 sourceNot checked — no full text of the source
• cites Raja, 2019Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites de Wit & Withagen, 2019 sourcePartly supported

  The opening passage quotes a criticism of ecological psychology for treating the brain as “wonder tissue” that resonates with affordances, and says neural contributions have generally received less attention. The Overview of the special issue also describes an organism–environment system as resonating. Neither passage says that resonance remains a metaphor or lacks a mechanistic account specifically. (LLM, conf=0.75)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing).

Falandays and colleagues (2021) proposed a homeostatic reservoir model with dynamic set-points, which they described as allostatic.

“Falandays et al. (2021) introduced a homeostatic reservoir model that avoids this criticism to some extent, using nodes that can be described as ''allostatic,'' meaning their homeostatic set-points are dynamic.”

• cites Falandays et al. (2021) sourceSupported

Classified as probably not a substantial claim.

Falandays and colleagues (2021) suggested that homeostasis at the neuronal level could underlie apparent predictive processing in the brain.

“Falandays et al. (2021) suggested that neuronal homeostasis may be one potential mechanism for apparent ''predictive processing'' in the brain.”

• cites Falandays et al. (2021) sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The reservoir network model was first introduced by Falandays et al. (2021) and applied to language processing.

“This model was first introduced by Falandays et al. (2021), where it was applied to the context of language processing.”

• cites Falandays et al. (2021) sourceSupported

May combine two claims.

The network regulated its own activity in a manner that followed the input's temporal dynamics, acting as a predictive model for control.

“In other words, the network controlled its own flow of activity in a way that tracked the temporal dynamics of the input, embodying a predictive model of the input sequence for the purposes of control.”

• cites Falandays et al. (2021) sourcePartly supported

  Section 5.2 says nodes must anticipate the temporal structure of inputs to stabilize their activity, supporting the claim about control that tracks input dynamics. But the Table 2 discussion calls this adaptation only functionally predictive: it says the network couples with its environment rather than representationally modeling it. Describing the network as embodying a predictive model overstates that distinction. (LLM, conf=0.86)

  Experimental — accuracy not yet measured

May combine two claims.

Falandays et al. observed that the reservoir generated endogenous activity that offset the input in real time, sending inhibition to sensory-receiving nodes and excitation to nodes requiring a boost.

“They found that the reservoir adapted to produce endogenous activity that compensated for the input in real time, routing inhibitory input to nodes that were receiving sensory inputs, and excitatory input to nodes that needed a boost.”

• cites Falandays et al. (2021) sourceSupported

Classified as probably not a substantial claim.

If homeostatic mechanisms are built into models of the CNS, organism-level control may emerge as a natural consequence.

“This position suggests that if we build homeostatic mechanisms directly into our models of the CNS, organism-level control may emerge as a natural consequence.”

• cites O'Leary & Wyllie, 2011 sourcePartly supported

  The paper’s J Physiol 589.20 section describes global control of the nervous system as emerging from a hierarchy of regulatory mechanisms, and ‘The need for control’ extends that hierarchy to entire organisms. It does not say that adding homeostatic mechanisms to CNS models will produce organism-level control; that modeling prediction is the citing authors’ extrapolation. (LLM, conf=0.84)

  Experimental — accuracy not yet measured

May combine two claims.

In that model, neurons regulate overall firing rates while lower-order set points can vary over time.

“In the model from Falandays et al. (2021), neurons pursue homeostasis at the level of overall firing rates, while permitting of variability over time in lower-order set points.”

• cites Falandays et al. (2021) sourcePartly supported

  The Model description and Network and Node Properties sections say that nodes pursue homeostasis by adjusting a target activation level that can vary over time. They do not describe the homeostatic target as an overall firing rate or identify the variable targets as lower-order set points; the model instead distinguishes activation from the threshold at which a node spikes. Page numbers were not available in our copy. (LLM, conf=0.85)

  Experimental — accuracy not yet measured

May combine two claims.

The model showed signs of predictive processing, including sequence completion and activity spikes to unexpected inputs, but did not use explicit predictions or prediction errors.

“As a result, this model exhibited some behavioral signatures of predictive processing, such as sequence completion and spikes of activity in response to unexpected inputs, but without the use of explicit predictions or prediction errors.”

• cites Falandays et al. (2021) sourceSupported

May combine two claims.

Reservoir networks with local homeostatic control resemble biological dynamics in important ways, including 1/f noise and neural avalanches.

“Thus, in addition to having desirable properties for a cognitive architecture, reservoir networks with local homeostatic control resemble biological dynamics in important ways.”

• cites Kello, 2013 sourceNot checked — no full text of the source
• cites Kello et al., 2011Supported
• cites Rodny et al., 2017 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Kagan et al. (2022) inspired the second case study; they trained a cortical tissue culture to play Pong.

“The second case study was inspired by recent work from Kagan et al. (2022), in which a culture of cortical tissue was trained to play Pong.”

• cites Kagan et al. (2022) sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Existing free-energy principle models of neural dynamics suggest that neural populations explicitly encode a generative model and perform Bayesian inference, which seems difficult to reconcile with the authors' model.

“Indeed, existing FEP models of neural dynamics suggest that neural populations explicitly encode a generative model and perform Bayesian inference (Ramstead et al., 2021), and in that respect, would seem difficult to reconcile with our model.”

• cites Ramstead et al., 2021 sourcePartly supported

  Section 5 (Discussion; page unavailable) says existing neural process theory models neuron or population activity as encoding posterior beliefs, while §2.4 describes neurons using a generative model for inference. But §2.2 characterizes the generative model as implicit and inference as something the dynamics look ‘as if’ they perform; the claim’s ‘explicitly encode’ formulation overstates the source. The apparent difficulty of reconciling those models is the citing authors’ own assessment. (LLM, conf=0.85)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing).

Ecological psychology and the free-energy principle have often been treated as competing frameworks because the former rejects representational notions and the latter relies on them extensively.

“Ecological psychology and the FEP have often been framed as competing frameworks for cognitive science (Bruineberg et al., 2022; though for an argument to the contrary, see Bruineberg et al., 2018), given that the former rejects representational notions, and the latter relies on them extensively.”

• cites Bruineberg et al., 2022Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Bruineberg et al., 2018 sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing).

The authors' interpretation of how homeostasis leads to behavioral control resembles the free-energy principle and active inference.

“Although we situate this work in the context of ecological psychology, it is worth pointing out that our interpretation of how the homeostatic mechanism leads to behavioral control is reminiscent of another theoretical paradigm, the free-energy principle (FEP; Friston, 2010b), and its related process theory, active inference (Ramstead et al., 2018).”

• cites Friston, 2010bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Ramstead et al., 2018 sourceSupported

Classified as probably not a substantial claim.

That culture achieved slightly better-than-chance Pong performance when hits were paired with predictable stimulation and misses with unpredictable stimulation.

“The culture learned to play Pong with slightly above-chance performance when it was trained by providing a predictable pattern of exogenous stimulation when the paddle hit the ball, or an unpredictable pattern of stimulation when the paddle missed the ball.”

• cites Kagan et al. (2022) sourcePartly supported

  STAR Methods (Method details; page unavailable) describes predictable stimulation after hits and unpredictable stimulation after misses. The Results (‘BNNs learn over time when embodied in a gameplay environment’; page unavailable) report that the cultures improved and outperformed random-noise controls. The paper does not establish a chance-performance baseline or characterize the advantage as ‘slightly’ above chance. (LLM, conf=0.84)

  Experimental — accuracy not yet measured

May combine two claims.

The model has previously produced patterns reminiscent of predictive processing in language, and in this work produced context-appropriate behavior in three distinct settings when embedded in an action-perception loop.

“This model has previously been shown to produce patterns reminiscent of predictive-processing when dealing with language (Falandays et al., 2021), and in this work has been shown capable of producing context-appropriate behavior in three distinct settings, when embedded within an action-perception loop.”

• cites Falandays et al., 2021 sourceSupported

May combine two claims.

The reservoir network avoids walls while controlling a mobile agent.

“This network spontaneously produces behaviors that are adaptive in each context, including (1) visually tracking a moving object, (2) substantially above-chance performance in the arcade game Pong, (2) and avoiding walls while controlling a mobile agent.”

• AssumptionRests on an assumption: The three agent-environment tasks are representative of contexts in which biological organisms must produce adaptive behavior.

May combine two claims.

If the stimulus reverses direction, the agent follows after a short lag and sometimes loses it.

“When the stimulus changes directions, the agent turns to follow it with a brief delay, occasionally losing track of the stimulus.”

May combine two claims.

Such activity spikes happen either at stimulus direction changes or when the agent loses the stimulus and then finds it again.

“Cross-referencing again with the agent-stimulus dynamics (Fig. 4A), it is apparent that these spikes in activity occur either when the stimulus changes direction (e.g.around t = 4200), or when the agent has lost track of the stimulus and encounters it again (e.g.around t = 3000).”

May combine two claims.

Across 500 runs, the average hit rate was 58.2% (SD = 9.95%), far exceeding the 20% chance level.

“The mean percentage of hits over runs was 58.2% (SD = 9.95%), well above the chance performance of 20%.”

May combine two claims.

Even with noise, the agent still tends to avoid walls and to seek circular movement patterns that are sometimes disrupted.

“Nonetheless, the agent maintains a tendency to avoid walls, and to seek out circular movement patterns that are occasionally disrupted.”

May combine two claims.

After a stable movement pattern appears, the network's dynamics are either fully stable or a limit cycle in which only a few nodes change values.

“Investigation of the network dynamics once a stable movement pattern has been discovered shows that this involves either completely stable activity, or a limit cycle (repeating pattern of spikes over time) with only a small number of nodes changing values.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In one representative run, the network reached a stable movement pattern and network state in roughly 300 time steps.

“The first column of Fig. 9 shows the first 1000 timesteps of a representative run in which the network discovered a stable movement pattern and network state within around 300 time steps.”

May combine two claims.

The authors have shown that adaptive behavior emerges spontaneously in three simulated scenarios: a rotating agent with a moving stimulus, Pong, and a mobile agent in a walled space.

“We have shown that in three distinct scenarios-(1) a rotating agent in an environment with a moving stimulus, (2) the game Pong, and (3) a mobile agent in a walled space-adaptive behavior spontaneously emerges.”

May combine two claims.

Signs of neural encoding appear only briefly; across longer spans, no consistent tuning pattern for individual nodes emerges.

“In sum, evidence for neural encodings can only be found over short time windows; when we observe the system for a longer period of time, we can find no stable pattern of tunings for particular nodes.”

May combine two claims.

Performance peaked almost from the start: the first 50 opportunities averaged 57.86% (SD = 10.5%) and the last 50 averaged 57.86% (SD = 12%).

“Comparing the likelihood of hits in the first 50 opportunities on each run (M = 57.86%,SD = 10.5%) to the last 50 opportunities (M = 57.86%,SD = 12%), we see that the model was at peak performance from near the beginning of a run (see Fig. 7, grayscale points).”

May combine two claims.

In the baseline and added-noise conditions, the agent hits walls several times in the first few hundred time steps and then finds a stable movement pattern with no further wall contact.

“In the baseline condition and the added sensory noise condition, Fig. 9 shows that the agent makes a number of collisions with the wall in the first few hundred time steps (black ticks beneath heatmaps), after which it discovers a stable movement pattern that results in no further contact with walls.”

May combine two claims.

In the present work, some nodes had strong correspondences with clockwise or counterclockwise movement, but these correspondences drifted throughout the reservoir over time.

“Similar outcomes have been observed here (see Fig. 5), with some nodes having strong correspondences with movement either clockwise or counterclockwise, but with these correspondences drifting throughout the reservoir over time.”

May combine two claims.

The network's adaptive behavior emerges from local homeostatic mechanisms at the node level and does not depend on stable or recurring activity patterns.

“demonstrating that apparently-adaptive behavior at the agent level emerges from the homeostatic mechanism at the level of nodes, and is not dependent upon the formation of stable and/or recurring activity patterns of the kind that might be expected within a representational theory of CNS.”

Probably not a substantial claim: may not pass S2 (checkable).

Supplying a mechanistic account of resonance requires a non-representational account of central nervous system dynamics.

“Supplying a mechanistic account of resonance requires a non-representational account of central nervous system (CNS) dynamics.”

May combine two claims.

Artificial neural network models rarely incorporate homeostatic principles; those that do often use evolutionary algorithms, leaving unclear how much adaptability is due to homeostasis versus evolved architecture.

“The few ANN models of which we are aware that have included homeostatic mechanisms (Di Paolo & Iizuka, 2008;Iizuka & Di Paolo, 2007) have used evolutionary algorithms to create viable architectures, leaving open the question of how much of their adaptability is due to homeostasis, and how much to the particular architecture that was evolved”

• cites Di Paolo & Iizuka, 2008 sourceNot checked — no full text of the source
• cites Iizuka & Di Paolo, 2007 sourceNot checked — no full text of the source

May combine two claims.

Actions that keep input flowing steadily into the network receive an implicit reward, whereas actions that disrupt input do not.

“In sum, behaviors that maintain a consistent flow of input to the network are implicitly rewarded, while behaviors that undermine the input to the network are not.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Understanding how self-organization of CNS dynamics links to self-organization of organism-environment dynamics is necessary for a mechanistic account of resonance.

“However, in order to develop a mechanistic account of the concept of resonance, it is necessary to understand how the self-organization of CNS dynamics is linked to the self-organization of organism-environment dynamics.”

May combine two claims.

During these episodes, a disturbance causes an activity spike, which in turn initiates homeostatic updating.

“At each of those events, the agent encounters a perturbation in the flow through the network, which leads to a spike in activity that triggers homeostatic updating.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

The network's activity need not be seen as representational from the system's own perspective, though it can be read out as representational by an outside observer.

“Third, and perhaps most importantly, we suggest that their activity need not be seen as representational from the perspective of the system itself, though it can be read out as representational to an outside observer.”

Classified as probably not a substantial claim. May combine two claims.

Brain activity and behavior are self-organizing systems.

“An important point in ecological psychology, missing from many models of the CNS, is that brain activity and behavior are self-organizing systems (Kelso, 1995).”

• cites Kelso, 1995Not checked — no full text of the source

May combine two claims.

The authors think Falandays et al.'s model can provide a non-representational explanation of CNS prediction, which could be valuable for ecological psychology.

“Importantly, we believe the model introduced by Falandays et al. (2021) allows one to cast a non-representational account of how the CNS ''predicts,'' making it potentially useful in the ecological framework.”

Classified as probably not a substantial claim. May combine two claims.

Reservoir networks satisfy three desiderata for cognitive theories: dynamic memory, integration across timescales, and multimodal integration.

“Dale and Kello (2018) point out that reservoir networks are also interesting as a model of cognition, because they satisfy three important desiderata for contemporary theories of cognition.”

• cites Dale and Kello (2018) sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Rodny et al. (2017) reported localist representations in critical branching networks that drifted, paralleling representational drift seen in mice (Rule et al., 2019).

“However, previous work by Rodny et al. (2017) found the presence of localist representations in critical branching networks that drifted over time, similar to demonstrations of representational drift in mice (Rule et al., 2019).”

• cites Rodny et al. (2017) sourceNot checked — no full text of the source
• cites Rule et al., 2019 sourceNot checked — no full text of the source

May combine two claims.

The network's behavior can be explained as multi-scale resonance, where individual nodes resonate to local energy flows, driving movement and new perceptions, allowing the agent to resonate to external information.

“We explain this as multi-scale resonance, whereby individual nodes resonate to flows of energy in their immediate environment, which in turn drives movement and new perceptions, ultimately allowing the agent as a whole to resonate to information in the external environment.”

Probably not a substantial claim: may not pass S4 (based).

When spiking activity is what neurons regulate, homeostatic and synaptic changes let the network adjust to repeated perturbations and return spiking toward a target.

“If spiking activity is the variable being regulated by neurons, neuronal homeostasis and synaptic updates allow the network to eventually adapt to the regular pattern of perturbations, bringing spiking activity back towards a target profile.”

• AssumptionThe authors state this as an assumption: Individual neurons require at least some ongoing input to survive.

May combine two claims.

Should the agent turn away from the stimulus, input to the network ceases, and the reservoir falls silent until the stimulus reappears.

“If, when this activity spreads to the effector layer, the agent turns in the opposite direction from the stimulus, activity will stop entering the network entirely, and the reservoir will eventually stop spiking until the stimulus comes back around (or the agent comes back around to the stimulus).”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

One potential fix is to discover a network state that controls action in a way that makes the input regular once more.

“One possible solution is to find a network state that regulates action so as to render the input regular once again.”

May combine two claims.

When tracking persists for a sufficient period, learning stabilizes and the behavior continues without end.

“If the agent tracks the stimulus for a sufficient amount of time, learning can stabilize and the ongoing behavior will be sustained indefinitely.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Intrinsic fluctuations poised at the edge of chaos are thought to maximize computational efficiency and flexibility to switch between adaptive regimes.

“Intrinsic fluctuations that are poised ''at the edge of chaos'' are thought to maximize the computational efficiency of such systems and the flexibility to switch between adaptive regimes (Bertschinger & Natschla ¨ger, 2004).”

• cites Bertschinger & Natschläger, 2004 sourceNot checked — no full text of the source

May combine two claims.

When fed inputs from a simple probabilistic grammar, the network exhibited behavioral signatures of predictive coding, including increased activity for surprising inputs and sequence completion, suggesting it can entrain to complex sensory patterns without supervision.

“When the network was fed inputs generated by a simple probabilistic grammar, it was shown to exhibit behavioral signatures associated with ''predictive coding,'' including increased activity for surprising inputs and sequence completion, suggesting that the network is able to entrain itself to complex sensory patterns that unfold over time, without the need for supervision.”

Probably not a substantial claim: may not pass S1 (asserted). May combine two claims.

A prominent view in cognitive science holds that the brain predicts upcoming sensory inputs and uses prediction errors to adjust an internal model of the environment.

“A prominent general view in cognitive science today is that the brain learns a model of the environment by predicting upcoming sensory inputs and using prediction errors to adjust parameters of the model (Hohwy, 2018).”

• cites Hohwy, 2018 sourceNot checked — no full text of the source

May combine two claims.

The wall-avoidance model provides clear evidence of learning, whereas the Pong case did not clearly show learning.

“Thus, while learning (in the sense of improvement in performance over time) was not clearly evident in the Pong case study (perhaps because the timescale of learning was much faster than the peformance measure) the wall-avoidance model shows clear evidence of learning.”

May combine two claims.

Following the ball might be implicitly reinforced because it helps stabilize the network, which in turn raises hit probability.

“As in the first case study, tracking the movement of the ball may be implicitly rewarded by virtue of facilitating stability within the network, leading to a higher likelihood of hits.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The free-energy principle holds that organisms act to avoid surprising states by optimizing a generative model of the environment and the consequences of actions.

“The FEP holds that organisms act so as to avoid ''surprising'' states, by optimizing a model of the environment and the consequences of actions (the ''generative model'').”

• cites Friston, 2010bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

In the wall-avoidance model, the random turn caused by hitting a wall is not a necessary feature of the mechanics, and without it a different, possibly non-adaptive behavior might emerge.

“For example, in our wall-avoidance model, the fact that hitting a wall produces a random turn is not a necessary feature of the mechanics, and perhaps without this feature, a different pattern of behavior would emerge that would not seem ''adaptive''”

Probably not a substantial claim: may not pass S3 (load-bearing).

Masumori et al. (2015) inspired the third case study.

“The final case study was inspired by work from Masumori et al. (2015).”

• cites Masumori et al. (2015) sourceNot checked — no full text of the source

May combine two claims.

Unpredictable input may drive parameter changes, causing the network to seek a configuration that makes stimulation predictable and thus reduces misses.

“Unpredictable patterns of stimuli may lead to adjustments of network parameters, leading the network to search the space of parameters until a set is discovered that renders the stimulation predictable, which will consequently minimize misses.”

Probably not a substantial claim: may not pass S1 (asserted), S4 (based).

Were the network able to make sensory input fully predictable, it would eventually stabilize its activity entirely and stop altering parameters.

“If this were the case, the network would eventually be able to completely stabilize its activity, and would cease changing parameters after a time.”

• AssumptionThe authors state this as an assumption: Individual neurons require at least some ongoing input to survive.

May combine two claims.

In scenario 3, hitting a wall causes a sudden turn and maximally unpredictable activity flows, while avoiding walls allows complete stability of sensory flow.

“In scenario 3, hitting a wall produces a sudden turn either clockwise or counter-clockwise.This creates a situation in which hitting walls generates maximally unpredictable flows of activity, whereas avoiding walls allows for complete stability of sensory flow.”

Classified as probably not a substantial claim. May combine two claims.

Masumori et al. cultured cortical cells on a high-density microelectrode array and used them to drive a mobile robot equipped with wall detectors.

“Similarly to the work by Kagan et al. (2022) discussed in the previous case study, Masumori et al. (2015) grew a culture of cortical cells on a high-density microelectrode array, which was used to control a mobile robot with sensors that detected the presence of walls.”

• cites Masumori et al. (2015) sourceNot checked — no full text of the source

Classified as probably not a substantial claim.

Raja and colleagues proposed Anderson's neural reuse hypothesis, which describes brain activity as transiently active local neuronal subsystems (TALoNS) that coordinate brain, body, and world.

“Raja and colleagues have suggested, as a foundation for this work, Anderson's neural reuse hypothesis, which casts brain activity in terms of transiently active local neuronal subsystems (TALoNS), which are ''task-specific neural synergies that coordinate brain, body, and world'' (Raja &”

• cites Raja & Anderson, 2019 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Natural selection must generate physiological constraints such that whatever activity flow keeps neurons alive also benefits the whole organism.

“In other words, natural selection must generate a set of physiological constraints such that whatever flow of activity keeps neurons alive is also good for the entire organism.”

Classified as probably not a substantial claim.

The authors contend that the models' simplifications are appropriate analogs of real constraints faced by organisms.

“Thus, while our models clearly contains many simplifications, we contend that these are appropriate analogs of real constraints faced by organisms.”

May combine two claims.

In scenario 2, missing the Pong ball resets its position and creates maximally unpredictable input, while hitting the ball preserves a continuous, predictably changing input trajectory.

“In scenario 2, missing the Pong ball leads to a sudden reset of the ball's position, again leading to maximally unpredictable input, whereas hitting the ball preserves a continuous trajectory of inputs that changes in a predictable manner.”

Classified as probably not a substantial claim.

To verify the agent responded to changing stimuli rather than always turning one way, the stimulus reversed direction every 720 steps.

“The stimulus was set to switch directions every 720 time steps, or two full rotations, in order to check that the agent was responsive to changing stimuli, rather than always rotating in one direction.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Forward movement typically produces a certain kind of optic flow, but this relation is interrupted if an organism hits a wall.

“For example, while forward movement will typically produce a certain kind of optic flow, this relation will be interrupted if an organism hits a wall.”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are possible scenarios in which the behavior that facilitates homeostasis would not be the behavior the authors consider adaptive.

“However, it should be noted that there are possible scenarios in which the behavior that facilitates homeostasis will not be the ''adaptive'' (per our expectations) behavior.”

May combine two claims.

If an organism relies on a stable relation between movement and optic flow for homeostasis, hitting a wall disrupts homeostasis and leads to a behavior change.

“If the organism was relying on a stable relation between movement and optic flow to achieve homeostasis, then hitting a wall will disrupt homeostasis, and therefore lead to a change in behavior.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

From an ecological-psychology-inspired view, natural selection's role is to produce organisms whose action-perception loops that facilitate homeostasis are precisely those that are adaptive.

“In an ecological-psychologyinspired view of evolution, we would suggest that it is the role of natural selection to produce organisms for which the action-perception loops that facilitate homeostasis are precisely those that are adaptive.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Because the model has no analog of fitness, calling its behavior adaptive is based on the authors' preconceived notions of what behaviors would be adaptive in a given context.

“it should be noted that, because our model has no analog of fitness, describing the behavior as ''adaptive'' is purely based on our preconceived notions for what behaviors would be adaptive in a given context”

May combine two claims.

As the ball nears the paddle, the input pattern shifts quickly, boosting activity across the network.

“Thus, the input pattern begins to change very rapidly when the ball is about to pass the paddle, which leads to an increase of activity throughout the network.”

• AssumptionRests on an assumption: Closer objects seem to move faster to an observer.

Probably not a substantial claim: may not pass S3 (load-bearing).

Kagan et al. read their results through the free-energy principle, proposing that the cells minimized prediction errors.

“The authors interpreted these findings through the lens of the free-energy principle (Friston, 2010a), suggesting that the cells learned to minimize prediction errors.”

• cites Friston, 2010aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

As a check on the explanation, the model was rerun using allocentric sensory encoding.

“To confirm the above explanation, we re-ran the model with an allocentric sensory encoding.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The free-energy principle is framed at a higher level of abstraction than the authors' model and is not committed to any particular mechanistic account of brain dynamics.

“However, the FEP is framed at a higher level of abstraction than our model, and is not committed to any particular mechanistic account of brain dynamics.”

Probably not a substantial claim: may not pass S2 (checkable).

The authors take the model's mechanics to be a reasonable analog of real conditions faced by biological organisms.

“While this criticism is accurate to an extent, we take these mechanics to be a reasonable analog of the real conditions faced by biological organisms.”

Classified as probably not a substantial claim. May combine two claims.

Raja's work has highlighted that ecological psychology lacks a mechanistic account of resonance, which would require a non-representational account of central nervous system activity.

“In recent years, work from Raja (2018Raja ( , 2019Raja ( , 2021) ) has called attention to the lack of a mechanistic account of the concept of ''resonance'' within ecological psychology, which requires a story about the CNS that does not fall back on a representationalist account of brain activity.”

• cites Raja (2018, 2019, 2021)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

That disturbance makes movement more likely, raising the chance the paddle meets the ball.

“This perturbation increases the likelihood of movement, thereby increasing the odds that the paddle will contact the ball (it could move in the wrong direction, but in this situation staying still means a miss, so erratic movement is better than nothing).”

• AssumptionRests on an assumption: Closer objects seem to move faster to an observer.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The authors see at least some overlap with the free-energy principle: predictable stimuli may make homeostasis easier, so a homeostatic system might act to make stimuli predictable.

“There may be important theoretical differences between our account and those associated with the free-energy principle, which is beyond the scope of this paper to discuss in detail, but there is at least some clear overlap: homeostasis may be more achievable when patterns of stimuli are predictable, therefore a system that pursues homeostasis may act so as to render stimuli predictable.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In a Pong game where the goal is to avoid the ball, the model would perform poorly because nothing pushes it out of the observed regime of tracking the ball.

“For example, if we imagine a game of Pong in which the goal is to avoid the ball, our model would do quite poorly, because there is nothing to push it out of the observed regime of tracking the ball.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Masumori et al. suggested that actions that stop stimulation let the network stabilize, while ongoing stimulation drives adaptation until a stable avoidance pattern emerges.

“The authors proposed that this result occurs because movements that lead to the cessation of stimulation (i.e.avoidance) allow the network to stabilize activity, whereas continued stimulation leads to adaptation in the network until a stable avoidant pattern is discovered.”

• cites Masumori et al. (2015) sourceNot checked — no full text of the source

Classified as probably not a substantial claim.

Masumori et al. observed that the cell culture spontaneously got better at avoiding walls, with no external reward.

“They found that the collection of cells spontaneously improved in its tendency to avoid walls, without the need for any external reward.”

• cites Masumori et al. (2015) sourceNot checked — no full text of the source

May combine two claims.

If the ball is headed for the paddle, the angle shifts less quickly, letting the network settle its activity and making it more likely the paddle stays put.

“Conversely, when the ball is on course to hit the paddle, the angle relative to the center changes more slowly (or maybe not at all), giving the network more opportunity to stabilize activity, and therefore increasing the odds that the paddle remains still.”

• AssumptionRests on an assumption: Closer objects seem to move faster to an observer.

May combine two claims.

The model's sensors encode the ball's angle with respect to the paddle, and this angle shifts faster as the ball approaches.

“In our model, the sensory array encodes the angle of the ball relative to the angle of the paddle, and that angle changes more rapidly (for a constant speed of the ball) as the ball gets closer to the paddle.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

A reservoir computer can be trained to match its own input, becoming an autoencoder, so these are not exclusive categories.

“Note that a reservoir computer can be trained to match its own input, becoming an autoencoder, so these are not exclusive categories.”

Probably not a substantial claim: may not pass S2 (checkable).

Homeostatic networks offer a promising path toward a mechanistic account of the ecological concept of resonance.

“In conclusion, we suggest that homeostatic networks offer a promising path towards providing a mechanistic account of the ecological concept of ''resonance.''”

Classified as probably not a substantial claim. May combine two claims.

The authors emphasize their proposal is a 'how-possibly' explanation: the model shows one possible mechanism for resonance, but further work is needed to determine if or to what extent it accounts for human or other organism behavior.

“We emphasize that our proposal is a form of ''how-possibly'' explanation(Dray, 1968): our model shows one possible mechanism by which resonance could occur, but much further work will be necessary to determine if, or to what extent, something like this mechanism actually accounts for the behavior of humans or other organisms.”

• cites Dray, 1968 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

The authors' examples point to a potential evolutionary 'hack': a head start on intelligence with minimal barrier to entry.

“Our examples point to a potential evolutionary ''hack''-a head start on intelligence with minimal barrier to entry.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

The model may be applicable to a wider variety of domains and may help explain many kinds of adaptive behavior without appealing to mental representation.

“This suggests that our model may be applicable to an even wider variety of domains, and may help shed light on the emergence of many kinds of adaptive behavior, without the need to appeal to mental representation.”

May combine two claims.

If there are no stably recurring patterns, there is no entity R0, so there is nothing to ascribe functions to or to construct a system of representations, and requirements R2 and R3 are rejected.

“In fact, if there are no stably recurring patterns at all, there is no entity that we can call R 0 , hence nothing to ascribe functions to, nor to construct a system of representations.Therefore, we can also reject R2 and R3 outright.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The behavior of the agent-environment system is similar to a two-vehicle Braitenberg system, and the circular behaviors are comparable to 'revolving type relative equilibria'.

“The behavior of this agent-environment system is similar to that of a two-vehicle Braitenberg system, studied in depth byHotton and Yoshimi (in press).In particular, the circular behaviors are comparable to what are there studied as ''revolving type relative equilibria''.”

• cites Hotton and Yoshimi (in press)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

The authors' description has much in common with active inference and free-energy minimization, but there are important distinctions beyond the scope of the article.

“What we are describing has much in common with the activeinference and free-energy minimizing approach.However, we take there to be important distinctions as well, which are beyond the scope of this article to unpack.”

Probably not a substantial claim: may not pass S3 (load-bearing).

This view is consistent with a developmental systems theory approach to evolution, which suggests natural selection works on the entire developmental trajectory rather than just shaping an adult phenotype.

“This view is consistent with a developmental systems theory approach to evolution, which suggests that natural selection works upon the entire developmental trajectory, rather than simply shaping an adult phenotype (Griffiths & Tabery, 2013).”

• cites Griffiths & Tabery, 2013 sourceNot checked — no full text of the source

May combine two claims.

The authors suggest their model satisfies both the nature and knowledge hypotheses, meaning the agents neither use representations nor can be helpfully described as such.

“In contrast, we suggest that our model satisfies both the nature and knowledge hypothesis, meaning that our agents neither use representations, nor can be helpfully described as such.”

May combine two claims.

Because the activity patterns between the sensory and effector layers are not stable, they are not standardized, so requirement R1 for representation is rejected.

“As such, while we do see activity patterns that stand between a producer (the sensory input layer) and a consumer (the effector layer), these patterns are not stable, hence not standardized, and we can reject R1.”