A discussion essay follow-up to “What Biology Came Back to INSAR” mapping the complexity, dynamics, and regulation research already underway — and asking where this science should convene.
Note: This blog post is a work in progress, hyper-links, framing, checks and additional references will be added, but for time-sensitive reasons posting as available. Please comment corrections and questions!
There is a body of work in autism research that does not have a clear home at conferences, meeting or roundtables. It is methodologically serious, empirically active, increasingly cohesive across labs, and demonstrably converging across continents and disciplines. It is also structurally hard to see in the dominant taxonomy of the field. At INSAR 2026, a systems-science paper on autism could be squeezed into Brain Function, Digital Health, Sensory/Motor, Translational Neuroscience, Coexisting Conditions, or Methodological and Ethical Issues. But it would arrive as a visitor in someone else’s category. There is no topic area that names systems biology of autism, nonlinear neural dynamics, allostatic regulation, interpersonal synchrony, movement-sensing dynamics, or multi-scale regulatory modeling as a first-class scientific object. Each of these has a journal home. None has a conference home that recognizes it on its own terms.
What follows is a map of six threads of work already underway — with anchor citations — and a discussion of what unites them. The point is not to argue that INSAR is doing something wrong. The point is to make visible that there is a coherent science here that needs somewhere to land. Whether that somewhere is INSAR expanded, or a parallel convening organized by another body, is the open question this essay closes on.
Thread 1: Nonlinear neural dynamics
The neural signatures of autism are not well captured by averages. They are captured by what the activity does over time — its transitions, its variability, its temporal structure. Malaia, Ahn, and Rubchinsky (2026, Frontiers in Neuroscience) lay out the current state of this work: autism spectrum disorder is characterized by specific temporal patterning of neural activity, with changes in network responsiveness on short timescales producing consequences for learning, habituation, and predictive processing on longer ones. The authors are explicit that the deficit-based single-model account is no longer empirically tenable, and that the field needs theoretical and experimental developments grounded in dynamical systems.
Earlier work in this lineage — EEG complexity measures, microstate analyses, criticality, nonlinear dynamics of resting-state activity — has accumulated steadily over two decades. What is new is the willingness to name the framework. The Malaia et al. paper is an opinion piece in a peer-reviewed journal arguing, in effect, that dynamical systems analysis is not optional for understanding autistic brain function. It is necessary. No INSAR topic area names this work as such.
Thread 2: Movement-sensing dynamics
In 2013, Elizabeth Torres and the late Anne Donnellan co-edited a Research Topic in Frontiers called “Autism: The Movement Perspective.” It argued that movement and sensing are inseparable, that autistic differences are visible at the kinematic level long before they are visible at the behavioral level, and that the field’s reliance on observational rating scales was missing the data. In 2025, Torres, Travers, Delafield-Butt, and Srinivasan returned with “Autism: the movement (sensing) perspective a decade later” (Frontiers in Integrative Neuroscience) — 28 articles, more than 100 authors, covering kinematic signatures of prospective motor control, biorhythms, gait and posture, oculomotor dynamics, dyadic motor agency, and digital phenotyping at the micro-movement scale.
What this body of work has established is that autistic motor signatures are detectable with high reliability using objective kinematic measurement — Cook et al. on atypical basic movement kinematics in autism (showing jerk minimization differences); Forti et al. and others on prospective motor control in autistic children; whole-body movement variability and entropy differences detectable during ordinary activity. Movement is sensing. The peripheral nervous system is not peripheral. Cognition is embodied and time-extended. The implication is that “Sensory, Motor, and Repetitive Behaviors and Interests” — INSAR’s relevant topic area — is conceptually the wrong container. It treats movement as behavior to be observed. The work treats movement as the measurable expression of neural dynamics.
Thread 3: Active inference and predictive processing
For roughly a decade, a Bayesian and active-inference literature has been building competing computational accounts of autistic cognition and perception. The early attenuated-prior accounts (Pellicano & Burr’s “Anything but Less Predictable”; Lawson, Rees, and Friston’s aberrant precision account) have given way to more nuanced models. Arthur, Vine, Buckingham, Brosnan, Wilson, and Harris (2023, PLOS Computational Biology) tested predictive coding theories of autism using formal active inference models applied to grip forces and eye movements, concluding that atypical encoding of precision and context-sensitive adjustment, rather than blanket prior attenuation, better explains autistic perception. Shi, Allenmark, Theisinger, and colleagues (2025, Biological Psychiatry: Global Open Science) propose an “atypical iterative prior updating” account based on Bayesian modeling of perceptual adaptation.
This is not a fringe literature. It is computational psychiatry done well, with formal models, falsifiable predictions, and converging empirical results. It is also closely tied to Karl Friston’s free energy framework, which positions autistic cognition not as deficit but as a particular configuration of precision-weighted prediction and prediction-error processing. The clinical and personal implications are profound. There is no INSAR category for it.
Thread 4: Interactional dynamics and synchrony
Damian Milton’s 2012 “double empathy problem” reframed autistic social difficulty as a relational mismatch rather than an individual deficit. Bolis, Balsters, Wenderoth, Becchio, and Schilbach (2017, Psychopathology) translated this into a Bayesian framework: the “dialectical misattunement hypothesis,” which models psychiatric conditions including autism as dynamic interpersonal mismatches across nervous systems, levels, and timescales. Bolis and colleagues (2021, SCAN) then showed empirically that interpersonal similarity of autistic traits predicts friendship quality — i.e., autistic-to-autistic interaction is not the impaired case, it is its own case. Bolis, Dumas, and Schilbach (2023) extended this further into what they call “collective psychophysiology” and “inter-personalized psychiatry.”
In parallel, the neurotype-matching and social-motor synchrony literatures show that autistic interaction is not globally impaired. Rapport, synchrony, and mutual attunement depend on partner, task, environment, and neurotype match — Crompton et al. on the rapport advantage in matched-neurotype interaction; the broader social motor synchrony literature on how dyadic timing emerges from interactional conditions rather than from individual capacity. The relevant unit of analysis is not the isolated autistic individual but the interacting system. And the biobehavioral synchrony tradition — Tronick on mutual regulation, Feldman on parent-infant biobehavioral synchrony, Beebe on dyadic interaction microanalysis, Schore on right-brain attachment processes — has been quantifying nervous-system-to-nervous-system regulation for forty years, with measurable physiological markers (heart rate coupling, cortisol synchrony, autonomic state matching).
This is biology. It is also intervention science — the active ingredient of relational therapeutic work is empirically traceable. And it sits across multiple INSAR topic areas (Social Cognition, Family/Caregiver, First-Person Perspectives, Methodological and Ethical Issues) without belonging to any of them as biology.
Thread 5: Allostasis and regulatory physiology
Sterling and Eyer’s allostasis framework — stability through change, anticipatory regulation — later developed through McEwen’s work on allostatic load and allostatic overload, has become foundational across stress biology, developmental psychobiology, and psychoneuroendocrinology since the late 1980s. Ganzel, Morris, and Wethington (2010, Psychological Review) integrated it with developmental neuroscience to show how allostatic processes shape brain development. The autism applications have followed: work on cortisol dynamics (Corbett and others), heart-rate variability and vagal regulation, sympathetic-parasympathetic coupling, sleep architecture, immune-stress crosstalk, and the broader question of how regulatory bandwidth — the dynamic range across which a system can return to coherence — varies across neurotypes.
Hogenkamp, Sanghavi, and Natri (2026, Autism in Adulthood) explicitly framed this as an emergent paradigm question: a meta-framework treating neurodevelopmental variation as architectural heterogeneity in regulatory biology rather than deviation from a norm. The paper sits inside an autism journal but draws on physiology, complexity science, and developmental biology in roughly equal measure. ISPNE (the International Society of Psychoneuroendocrinology) hosts the stress-physiology work. Developmental psychobiology meetings host the regulatory caregiving work. INSAR has no place where a stress-and-regulation paper on autism would naturally land except by being cross-listed under something else.
Thread 6: Network and multi-system biology
The systems-biology literature on autism has, until recently, been dominated by molecular network analyses — protein-protein interaction networks, gene regulatory networks, GWAS-derived risk variant convergence on shared pathways. Randolph-Gips and Srinivasan (2012, Journal of Clinical Bioinformatics) made an early case that autism is a problem of physiological scales, arguing that systems-biology modeling would need to integrate metabolic, mitochondrial, immune, neurological, and behavioral data over time. More recent work has continued in this vein: dynamic perturbation propagation in autism gene networks; hypergraph analyses revealing regulatory hub structures; systems-biology gene prioritization integrating CNV data with network topology.
Alongside this, multi-system phenotyping has accumulated evidence that autism is rarely a brain-only condition. Co-occurring metabolic, immune, gastrointestinal, mitochondrial, and connective-tissue patterns appear with high frequency. Most of this work is published in molecular biology or systems biology journals, not autism-specific ones. INSAR’s “Genetics & Genomics” and “Molecular and Cellular Biology” categories are oriented toward etiology rather than systems integration. The translational neuroscience category is oriented toward bench-to-clinic, not multi-scale modeling. The work happens. The convening doesn’t.
What unites these threads
Across these six threads, there is now a substantial literature, appearing in journals from PLOS Computational Biology and Frontiers to Autism in Adulthood and broader neuroscience, developmental, and systems-biology venues. They look like different fields. They are not.
What unites them is a shared structural commitment. Each thread treats the phenomenon of interest as a dynamic, multi-scale, regulatory system rather than as a deficit signature in a linear pipeline. Each treats variability as informative rather than as noise. Each treats temporal structure as constitutive rather than secondary. Each treats the boundary between brain, body, and environment as porous rather than fixed. Each treats heterogeneity as architecture rather than as a measurement problem.
The technical name for this stance is complexity science applied to biology — sometimes called systems biology in the molecular sense, sometimes integrative physiology, sometimes dynamical systems development, sometimes network medicine, sometimes computational psychiatry. The autism-specific work above is, methodologically, all of these at once. It is doing the same kind of work that systems cardiology, systems immunology, and systems oncology have been doing for two decades — modeling the whole, in time, with appropriate mathematical tools, refusing to pretend the parts can be understood in isolation.
What it lacks is convening. The cardiology version has Gordon Conferences and dedicated network medicine meetings. The immunology version has Keystone symposia and dedicated systems immunology workshops. The autism version has nothing of comparable scale or programmatic identity.
What an interdisciplinary autism science could host
A meeting designed for this work would not be larger than INSAR; it might in fact be considerably smaller. What it would have is categorical fit. It would have programmatic tracks on:
- Multi-scale dynamics of neurodevelopmental variation
- Active inference and computational phenotyping
- Movement, sensing, and kinematic dynamics
- Interpersonal regulation, synchrony, and dyadic dynamics
- Allostasis, regulatory bandwidth, and stress physiology in development
- Network and multi-system biology of co-occurring conditions
- Methodological foundations: nonlinear time-series, network inference, multi-scale modeling
- Theoretical synthesis: complexity, predictive processing, ecological-enactive frameworks
- Lived experience as scientific data within a complexity framework
It would treat autistic researchers working in any of these tracks as biologists, computational scientists, or systems researchers, depending on their methods. It would not segregate “lived experience” as a separate axis to be balanced. It would have first-author and corresponding-author opportunities for autistic researchers across the program because their work happens to be where the field is going methodologically.
Such a meeting would also draw trainees who currently go elsewhere. Computational and systems neuroscience trainees who attend COSYNE because INSAR does not have categories for their work. Developmental psychobiology trainees who attend ISDP or ICIS. Psychoneuroendocrinology trainees who attend ISPNE. Active-inference researchers who attend the Active Inference Symposium. Network science researchers who attend NetSci. The audience exists. It is currently scattered. Convening it is a question of organization, not of demand.
A call, and an open invitation
There are two paths, and they are not mutually exclusive.
The first is for INSAR itself to take up this work. The 2026 meeting’s panel guidance already prioritizes interdisciplinary submissions and explicitly invites cross-cutting panels. The taxonomy could be expanded — three or four new topic areas covering nonlinear dynamics, multi-scale modeling, interpersonal regulation, and allostatic physiology would be a meaningful structural change. Master classes on complexity methods alongside the existing genetics and epidemiology offerings would draw the trainees the conference is concerned about losing. Plenaries on the integration of computational and lived-experience frameworks would model what the field could become. The 2027 and 2028 scientific program chairs have a real opportunity here, and the next generation of researchers — many of whom are autistic, many of whom are trained in complexity methods, many of whom would attend a meeting that had room for their work — is the constituency.
The second is for a parallel convening to emerge. A smaller, focused, single-track or limited-track meeting whose programmatic identity is systems and complexity science applied to neurodevelopmental variation and human regulation. Co-located with COSYNE or with ISPNE or as a Gordon Conference. Anchored by an organizing committee that includes autistic researchers, complexity scientists, allostasis researchers, computational psychiatrists, and dynamical movement scientists. Without competing with INSAR for scale or breadth — but offering a home for the work that INSAR’s current taxonomy cannot hold.
These two paths are complementary. A successful parallel convening would, over time, push INSAR toward expansion. A successful INSAR expansion would create the pipeline that makes the parallel convening viable at scale. The autism research community would, in either case, be better served — with more rigorous methods, more coherent frameworks, more integrated science, and more pathways for the trainees and researchers who are currently doing exceptional work without a conference home.
This is not a critique of INSAR. INSAR is the only autism-organized scientific convening of meaningful global scale, and it has done immense work building that community. This is an observation about the shape of the science that is emerging — work that is happening at the boundaries of existing disciplines, that has methodological coherence, and that needs convening.
If INSAR takes up the call, the field is well served. If another body builds a parallel home, the field is also well served. If both happen, the field is best served.
The work exists. The researchers exist. The convening, so far, does not. That is a solvable problem.
[1]: This essay draws on the framework of the Center for Adaptive Stress (CAS) and the Evolutionary Stress Framework, articulated in Hogenkamp, Sanghavi & Natri (2026), “Toward an Emergent Paradigm for Neurodiversity and Health,” Autism in Adulthood, DOI: 10.1177/25739581261433443. ESF is a conceptual lens; the framing of “regulatory bandwidth,” “architectural heterogeneity,” and “cooperative neurodiversity” used implicitly throughout this essay derives from that paper and the broader ESF publication arc. This essay is not medical advice.
Leave a Reply