The framework, called PERSCO, was developed by Sanne van Waveren, Zulfiqar Zaidi, and Matthew Gombolay at the Georgia Institute of Technology. Its central insight is that perceived safety can be treated as a tunable control problem rather than a fixed design choice. The researchers build on control barrier functions, or CBFs, mathematical constructs that guarantee a robot stays within a set of safe states by constraining its control inputs at every time step. Traditional CBFs enforce physical safety with static parameters that never change during an interaction. PERSCO instead parameterizes the CBF with two variables that directly shape how the robot’s behavior feels to nearby humans: the minimum distance the robot must keep from each person, and the maximum deceleration it is allowed to use, which in turn caps how fast it may approach anyone.

These two parameters have intuitive physical meaning. The distance parameter defines an intimate space around each person that the robot may never enter. The deceleration parameter determines the stopping distance the robot must be able to achieve at its current speed; a robot permitted stronger braking can safely travel faster, because it can halt in a shorter distance. By adjusting the pair, the controller can make the robot behave anywhere from maximally cautious to maximally assertive. The key question is which combination a given person actually perceives as safe, and that is something no designer can hard-code in advance, because perceptions vary widely between individuals and contexts.

To answer it, PERSCO treats the problem as active learning. Each person carries a hidden perceived safety function that maps any parameter pair to a judgment of safe or unsafe. The robot cannot observe this function directly, so it maintains a surrogate model, a classifier trained on feedback, and probes the boundary between safe and unsafe parameter regions. Crucially, the researchers designed the feedback to be as unobtrusive as possible. Rather than asking people to fill out Likert scales mid-task, PERSCO adopts a principle of perceived safe until proven unsafe: humans only signal when they feel uncomfortable, using a simple visual cue, in this study a handheld AprilTag sign raised toward a camera. Silence is treated as implicit safe feedback, provided the robot has logged enough close encounters with that person without any complaint.

The learning algorithm is engineered to minimize how often people must intervene. When unsafe feedback arrives, the robot updates its classifier and selects the next candidate parameters using a novel sampling strategy that balances two criteria: entropy, which targets regions where the model is most uncertain about where the boundary lies, and diversity, which favors candidates far from previously tested ones. Importantly, the sampler only considers parameters the model predicts to be safe, so the robot never deliberately behaves in a way it believes will alarm the human. When no unsafe feedback arrives after repeated close encounters, the robot gradually relaxes its parameters, stepping toward the least restrictive pair on the safety boundary, which maximizes task efficiency while remaining at the edge of what the person tolerates.

All of this runs inside a model predictive control loop with a 0.1-second time step, where the perceived safety CBF is enforced over the entire planning horizon while accounting for predicted human motion. A separate, unchanging physical safety CBF with the most aggressive parameters guarantees collision avoidance at all times, so no matter how the learned parameters evolve, the robot can always brake to a standstill before reaching anyone. When parameter updates suddenly tighten the constraints and the robot temporarily finds itself outside the new safe set, a gradual recovery strategy using a slack variable steers it back smoothly; in simulation this reduced jerk by 25 percent and angular acceleration by a factor of 3.6 compared to abrupt corrections, avoiding the jarring sidesteps that quick recovery methods would produce.

Simulation experiments validated the technical choices. Among three candidate classifiers, a support vector classifier with a radial basis function kernel proved the clear winner, updating in about 1.3 milliseconds on average, fast enough for real-time control, while the neural network and Gaussian process alternatives exceeded the control loop’s time budget. Against a battery of classical active learning baselines and black-box optimizers, including multi-armed bandits and Bayesian optimization, PERSCO sampling achieved high accuracy in recovering ground-truth safety parameters while producing the lowest ratio of unsafe feedback events. In a simulated workplace with three moving pedestrians, the system converged to near-optimal parameters in roughly eight minutes, both with and without noise injected into the feedback.

The decisive test came with real humans. Fifty-four participants, organized into eighteen groups of three, performed a workplace-inspired assembly task, walking between workstations to place LED pins on breadboards while a Boston Dynamics Spot robot navigated the space autonomously, covering 7,074 meters over the course of the study. Participants experienced three conditions: individual adaptation, in which the robot learned separate parameters for each person; collective adaptation, in which one shared parameter set was updated from anyone’s feedback; and an adversarial condition, in which the robot responded to feedback by becoming more aggressive rather than more cautious. The adversarial condition served as a control to test whether adaptation itself, or only feedback-aligned adaptation, improves how safe people feel.

The results were striking. Both aligned conditions significantly outperformed the adversarial one on perceived safety, comfort, and anxiety, all with p-values below .001 and large effect sizes. Collective adaptation scored highest overall, and participants raised their feedback signs significantly less often under collective updates, suggesting that people benefit from feedback provided by their teammates. A mediation analysis revealed that the effect of condition on perceived safety was fully mediated by the average size of the robot’s safety boundary, meaning the psychological benefit flowed directly from the geometric changes in the robot’s enforced constraints. Notably, individual adaptation offered no task-performance advantage over collective adaptation, contrary to the researchers’ hypothesis, possibly because fewer parameter changes allowed the robot to plan more consistently.

The authors are candid about limitations: the study took place in a controlled environment, sessions were capped at twelve minutes so parameters did not always converge, and treating silence as safe feedback assumes people are attentive enough to complain when they feel threatened. Still, the work marks a meaningful shift in how roboticists think about safety. By framing perceived safety as a quantity that can be measured, learned, and optimized alongside task performance, PERSCO argues that true safety encompasses psychological well-being, not just the absence of collisions. As robots move into warehouses, hospitals, and factories, the systems that earn human trust may be the ones that ask, in effect, how their presence feels, and adjust accordingly.

Subject of Research: Perceived-safe control of mobile robots using active learning from human feedback

Article Title: PERSCO: Perceived safe control of mobile robots in human groups with active learning

Article References: van Waveren, S., Zaidi, Z., & Gombolay, M. (2026). PERSCO: Perceived safe control of mobile robots in human groups with active learning. Autonomous Robots, 50(4), Article 43. https://doi.org/10.1007/s10514-026-10262-7

Image Credits: AI Generated

DOI: 10.1007/s10514-026-10262-7

Keywords: perceived safety, human-robot interaction, control barrier functions, active learning, mobile robots, model predictive control, social robotics, human-aware motion planning, adaptive control, Boston Dynamics Spot, user study, machine learning

The analysis examines the projects BabyRobot, CARER-AID, DE-ENIGMA, DREAM, PlusMe and IM-TWIN, all of which developed semi-autonomous robots or robotic soft toys intended to support autism interventions. Several used well-known humanoid platforms such as Nao, Kaspar and Probo, while the sister projects PlusMe and IM-TWIN designed toy-like soft devices described as Transitional Wearable Companions. None of the robots were intended to operate fully autonomously; many experiments relied on Wizard-of-Oz control, in which human operators quietly steered the machines while researchers worked on the technical scaffolding for semi-autonomous behavior. The interventions targeted socio-emotional and communicative skills such as joint attention, emotion recognition, perspective-taking and imitation, drawing heavily on behaviorist psychology, sometimes blended with developmental and cognitive approaches.

Because robotics papers rarely report ethics practices in detail, a problem the author likens to the ethics black-box known in artificial intelligence research, the study mined formal EU project deliverables, documents that grant agreements require researchers to produce for transparency and accountability. Using a critical thematic analysis informed by critical disability studies and critical autism studies, the author coded each project independently and then compared them across cases, looking explicitly for absences: ethical issues that were never mentioned, mentioned only vaguely, or discussed without nuance. Notably, the author writes from an insider standpoint as an autistic researcher, an approach grounded in what has been called a counterventionist strategy, which seeks to disrupt the normalization agenda embedded in human-computer interaction research rather than treat disability as a problem awaiting technological repair.

The projects varied in their ethical engagement. All obtained ethics committee approvals and parental consent, and all addressed privacy, data protection and physical safety to varying degrees. The DREAM project stood out for surveying stakeholders on whether robots should even be used in autism therapy, on emotional attachment and deception, and on robot autonomy. DE-ENIGMA conducted an extensive user-requirements analysis with educators, flagging concerns that children might become attached to robots, dependent on them, or deceived about the machines’ capacity for feeling and reciprocation, and it openly acknowledged sensory sensitivities. Yet only two projects, DE-ENIGMA and CARER-AID, explicitly sought the children’s assent rather than relying solely on parents. Several projects, including CARER-AID, PlusMe and IM-TWIN, never discussed children’s well-being at all, and the ethics deliverables of PlusMe and IM-TWIN were classified as confidential, illustrating how opaque the ethics of even publicly funded research can become.

The analysis then confronts a more uncomfortable question: whether the interventions themselves rest on solid evidence. Drawing on meta-research, the article notes that autism intervention science has produced roughly a thousand reports over three decades yet still lacks sufficient randomized controlled trials, and that a major meta-analysis of 150 group-based studies involving young autistic children found widespread risk of bias. When studies at significant risk of bias were removed, effect sizes dropped to statistical non-significance. Reported gains tend to be short-term, context-bound and poor at generalizing to everyday life, and adverse events are rarely monitored or reported. Widely recommended interventions, the review argues, therefore carry uncertain efficacy and unacknowledged potential harms, a finding that matters enormously when robots are presented as efficient new vehicles for delivering those same interventions.

The theoretical commitments embedded in the projects come in for particular scrutiny. Several projects invoked the Theory of Mind account of autism, which frames autistic people as lacking the ability to ascribe mental states to others, and some drew on the Empathizing-Systematizing framework, which characterizes the autistic brain as hyper-male and low in empathy. The review argues these models misrepresent and pathologize autistic people as a neurominority, and it contrasts them with the double empathy problem, formulated by autistic scholar Damian Milton, which holds that communication breakdowns between autistic and non-autistic people are mutual rather than a deficit located in one person. The DREAM project built its robot-assisted intervention on applied behavior analysis, or ABA, an operant-conditioning regimen of reinforcement and punishment that can run twenty to forty hours per week, described by critics as essentially a full-time job for a three-year-old.

The evidence on ABA’s safety is deeply contested. Researchers have documented that camouflaging autistic traits is a unique risk marker for suicidality, and mixed-methods studies have reported associations between ABA exposure and symptoms of post-traumatic stress, with autistic adults describing experiences of forced compliance, erosion of agency and trauma. A recent analysis of more than 17,000 autistic youth in a US national sample found preliminary evidence linking ABA exposure to increased mental health hospitalizations, though causation cannot be established. From a bioethics standpoint, it has been argued that ABA systematically violates the principles of nonmaleficence and justice by infringing on children’s autonomy. The review’s central charge is stark: the DREAM project did not mention this critique, and none of the projects employing behaviorist techniques acknowledged the weak evidence base or potential risks of the interventions their robots were designed to deliver.

The analysis also documents how the projects classified autistic children’s autonomy in pathologizing terms. In DE-ENIGMA, a child who failed to respond was labeled non-compliant, while needing several repetitions to comply counted as low engagement; self-regulatory behaviors arising from overstimulation were treated as interruptions to be corrected. In DREAM, distraction or task avoidance was categorized as maladaptive behavior. Even positive reinforcement carried costs: in CARER-AID, when institutionalized, intellectually disabled autistic children answered incorrectly, the Nao robot emitted an unpleasant sound with red eyes, a punishment technique the researchers defended as pedagogically important. In the sister projects, soft wearable companions were deliberately designed to arouse emotional attachment and reward eye contact and physical touch, despite evidence that eye contact can produce amygdala-mediated hyperarousal and genuine pain for many autistic people, and that joint attention in naturalistic play develops through non-gaze-based pathways in both autistic and neurotypical children.

Equally troubling is the pattern the review identifies as testimonial quieting, a structural form of epistemic injustice in which autistic people are deemed incapable of contributing knowledge about their own lives. Most projects justified excluding autistic children from design and feedback by citing communication difficulties, then consulted teachers, therapists and parents instead. The DREAM project’s stakeholder survey had no category for autistic respondents at all, and its collaborative-ethics approach, while democratic, drew overwhelmingly non-autistic input, a configuration critics warn can produce the tyranny of the majority for minority communities. Community research priority studies cited in the review show autistic people rank social-skills interventions among their lowest research priorities, while the primary barrier to socializing, according to experimental work on thin-slice judgments, is the reluctance of neurotypical peers to interact with them.

The article’s conclusion is that these technologies, as currently conceived, function less as assistive tools than as curative ones, in the sense defined by bioethicist Joseph Stramondo: any technology that treats disability as a problem to be erased through normalization is curative, whereas assistive technology expands access while preserving identity. Because the projects consistently framed autism as a costly public-health burden and defined success as neurotypical performance, their robots operated as disciplinary apparatuses that surveil, datafy and condition autistic behavior, in tension with the United Nations Convention on the Rights of Persons with Disabilities. The author argues that no ethically acceptable robot-assisted therapy for autistic children can exist until the field shifts from deficit-based models toward neurodiversity-affirming, participatory practices, in which autistic children and adults are treated as knowers and experts, consent and assent are continuously renegotiated, and the purpose of support becomes self-determination rather than compliance. Until roboticists ask whose values their technologies encode and whether a robot is even the right solution, the answer to the question posed by the paper’s title remains, uncomfortably, no.

Subject of Research: Ethics of robot-assisted autism interventions for autistic children

Article Title: Can there really be ethical robots for autistic children?

Article References: Elmadagli, C. (2026). Can there really be ethical robots for autistic children?. AI & SOCIETY. https://doi.org/10.1007/s00146-026-03372-4

Image Credits: AI Generated

DOI: 10.1007/s00146-026-03372-4

Keywords: social robotics, autism, robot-assisted therapy, research ethics, neurodiversity, technoableism, applied behavior analysis, Horizon 2020, epistemic injustice, human-robot interaction, critical autism studies, AI ethics

Horiuchi’s argument rests on a theory-driven synthesis that pulls together strands of research usually kept in separate silos: gerontechnology, social and assistive robotics, smart-home monitoring, the emerging scholarship on digital ageism, mainstream AI ethics, human enhancement debates, and capability-oriented accounts of aging and disability drawn from the tradition of Amartya Sen and Martha Nussbaum. From this interdisciplinary weave he extracts a four-pillar evaluative framework, comprising substantive opportunity, practical control, relation, and responsibility. Each pillar is a lens through which any AI- or robot-mediated care arrangement can be interrogated, and together they are meant to shift the conversation from deficit management toward what the author calls aging with dignity, relation, and room for revision and refusal.

The first pillar, substantive opportunity, borrows directly from the capability approach. It asks not whether a technology functions, but whether it widens the set of meaningful things a person can actually do and be. A fall-detection system that silently renders an apartment legible to remote monitors may reduce mortality statistics while doing nothing to restore a person’s ability to walk to the market, host friends, or pursue a lifelong hobby. Research on smart homes and home health monitoring, including systematic reviews of sensor technologies designed to support aging in place, has documented impressive technical reliability, but Horiuchi’s framework insists that reliability is not the same as opportunity. The question becomes: after the algorithm has done its work, is the older person’s real freedom larger or smaller?

The second pillar, practical control, tackles the everyday texture of human-machine interaction. Older adults are frequently positioned as passive beneficiaries of ambient intelligence, their homes instrumented with passive monitoring systems that require no action on their part. Ethnographic work on passive monitoring in low-income independent living has shown that residents are not simply grateful recipients; they actively interpret, negotiate, and sometimes resist the sense of being watched. Practical control asks who can turn the system off, who can correct its inferences, who understands what it is reporting, and whether the person retains what one might call breathing room in monitored space. A robot or monitoring platform that cannot be questioned, adjusted, or refused by its user, however technically elegant, fails this pillar.

The third pillar, relation, may be the most culturally charged. Socially assistive robots are already being trialed in dementia care, and studies of robots in advanced dementia care units report engagement and calming effects. Companion machines designed with trauma-informed care principles and humanoid robots used as memory trainers for people with mild cognitive impairment illustrate how far the field has moved. But Horiuchi, drawing on scholarship that explores affect and relationality across sites of intelligence and care, warns that a machine’s ability to occupy a relational slot does not automatically mean it enriches the human web of relations around the older person. The ethical question is whether a robot companion supplements fragile social networks or quietly substitutes for them, giving families and institutions a technological alibi for withdrawal.

The fourth pillar, responsibility, confronts the sprawling supply chains of contemporary AI. Large language models and generative systems in healthcare have prompted calls for urgent regulatory oversight, and research on the AI supply chain has revealed how modularity dislocates accountability, leaving developers with fragmented notions of who answers when something goes wrong. In eldercare, where misclassification can mean an overlooked fall or a false alarm of decline, this fragmentation is not an abstraction. Horiuchi argues that the ethical quality of a technology depends on the wider sociotechnical arrangement in which it is designed, deployed, maintained, interpreted, and refused, and that responsibility must therefore be assessed across that whole arrangement rather than pinned on a single device or codebase.

Two of the paper’s broader theses give the framework its intellectual bite. The first is that in super-aged societies, AI and robotics do not merely serve pre-existing needs; they help constitute what agency, dependency, care, and responsibility mean. This is the co-constitution thesis familiar from science and technology studies, where scholars have argued that aging and technology are mutually shaped rather than sequentially matched. A society that builds its care infrastructure around risk-minimizing sensors will produce different kinds of old age, and different kinds of dependency, than one that builds around mobility, creativity, and connection. The second thesis redefines enhancement itself. Against transhumanist framings that equate enhancement with youthfulness or optimization, Horiuchi argues that enhancement in later life should mean widening meaningful possibilities under changing bodily and social conditions, an idea that resonates with psychological models of successful aging built on selective optimization with compensation and with disability scholarship on the extended body.

The framework arrives at a moment of genuine technological ferment. Systematic reviews document a rapid proliferation of AI approaches for predicting and detecting cognitive decline, brain-computer interfaces being explored for cognitive enhancement in older people, personalized cognitive training platforms, and conversational agents whose perceptions and needs among older adults are only beginning to be mapped. Japan, the archetypal super-aged society, has poured public resources into eldercare automation, yet ethnographies of Japanese care robots have punctured boosterism by showing how robots reshape care work in ways that neither replace nor relieve human caregivers straightforwardly. Horiuchi’s contribution is to give researchers, care institutions, and policymakers a shared vocabulary for asking whether any of this machinery, at any point in its life cycle, is enabling or constraining the people it claims to serve.

Why does this reframing matter now? Because the default evaluative logic of the field is deficit management: count the falls prevented, the hospitalizations avoided, the minutes of caregiver time saved. Those metrics are not worthless, but they systematically undervalue what older people themselves often rank highest, namely autonomy, connection, and the ability to revise the terms of one’s own life. The capability-oriented literature on healthy aging has pushed in the same direction, shifting the lens from successful aging, which risks exalting a narrow standard of vigor, toward wellbeing through capability and human development for all older people, including those living with disability or frailty. Horiuchi’s four pillars give that philosophical shift a concrete evaluative instrument: for any AI-mediated arrangement, ask what substantive opportunities it opens, what practical control it grants, what relations it sustains or erodes, and where responsibility for its operation genuinely resides.

The paper also converges with a growing critical literature on AI ageism, which documents how algorithms trained on younger populations can misread older bodies and voices, and how digitalized services can exclude those least able to navigate them. By naming relation and responsibility as co-equal pillars alongside opportunity and control, the framework implicitly demands that ageism audits become part of standard deployment practice. And by insisting that refusal, the right to decline a technology altogether, is built into the ethical arrangement, it counters the quiet coercion that can seep into institutional care, where residents may feel they cannot say no to a monitor or a robot without seeming to reject care itself. In an era when longevity science is inflating expectations about radical life extension, Horiuchi’s modest, grounded proposal is a corrective: the measure of a technology for aging is not whether it defeats time, but whether it lets people fill the time they have with more of what matters to them.

Subject of Research: A conceptual framework for evaluating AI, robotics, and human enhancement by their effects on agency in aging societies

Article Title: Enabling later life: a four-pillar framework for AI, robotics, and human enhancement in super-aged societies

Article References: Horiuchi, S. (2026). Enabling later life: a four-pillar framework for AI, robotics, and human enhancement in super-aged societies. AI & SOCIETY. https://doi.org/10.1007/s00146-026-03381-3

Image Credits: AI Generated

DOI: 10.1007/s00146-026-03381-3

Keywords: artificial intelligence, robotics, aging, gerontechnology, human enhancement, digital ageism, care ethics, capability approach, social robotics, smart homes, AI ethics, super-aged societies

Autism is a developmental condition that emerges in early childhood and persists across the lifespan, profoundly shaping social behavior and often making the acquisition of learning and social skills more difficult. According to data cited by the U.S. Centers for Disease Control and Prevention, autism spectrum disorder now affects a substantial and growing share of children worldwide, which has intensified the search for educational tools that can supplement traditional teaching. Over the past two decades, interactive technologies—from computer-based programs to tablet applications—have been explored as supports for autistic learners, and social robots in particular have attracted intense interest from researchers and clinicians alike.

The logic behind robot-assisted intervention rests on a striking observation: many children with autism engage more readily with machines than with people. Robots are predictable, their expressions are simplified, and their behavior is consistent and rule-governed. For children who find the rapid, ambiguous signals of human social interaction overwhelming, a robot can act as a social intermediary—demanding enough to practice attention and turn-taking, yet simple enough to feel safe. Numerous studies on autism intervention have highlighted the effectiveness of social robots in behavioral treatments, including work showing that the NAO robot can improve eye-gaze attention in children with high-functioning autism, and that long-term child-robot interaction can sustain attention and engagement over repeated sessions.

What has been missing, the researchers argue, is evidence from authentic classroom settings. Most robot-assisted autism studies have taken place in laboratories or clinics, in structured dyadic interactions between a single child and a single robot. But learning in schools is inherently social and collective: children must share a teacher’s attention, follow group instructions, and navigate the presence of peers. Reviews of field-based studies of social robots in classrooms confirm that genuine classroom integration remains sparse. The new study was designed to begin filling that gap by placing the NAO robot directly into a group teaching context at a special education school, with human teachers and the robot working side by side.

The experimental design was deliberately collaborative. Rather than replacing the teacher, the NAO robot functioned as a co-facilitator of classroom activities. Special education teachers led sessions in partnership with the robot, creating a triadic learning environment in which interactions flowed among teacher, robot, and students. This arrangement reflects principles from established autism education frameworks such as TEACCH, which emphasizes structured teaching, and applied behavior analysis, which underpins many technology-mediated interventions. By distributing instructional roles between a sensitive human professional and a predictable, engaging machine, the researchers sought to foster a dynamic learning environment that neither agent could create alone. The study, conducted at Changshu Special Education School, was explicitly framed as a foundational investigation—a proof of concept in anticipation of extended robot-assisted classroom sessions to be introduced at a later date.

The technical appeal of the NAO platform is central to the study. NAO is a 58-centimeter-tall humanoid robot developed by SoftBank Robotics, equipped with cameras, microphones, tactile sensors, and articulated limbs that allow it to gesture, dance, speak, and emulate human movement. Its child-sized stature and expressive but simplified face reduce the perceptual complexity that often challenges children with autism. Research on gaze perception has shown that decoding gaze direction from combined head and eye rotations is an integrative challenge that differs in autistic individuals; NAO’s exaggerated, unambiguous head turns and eye movements sidestep much of that ambiguity, making it an ideal cueing device for directing attention toward learning materials. Prior work has also demonstrated that robots can reduce delays in gesture production among preschoolers with autism, suggesting that the platform’s motor expressiveness carries direct pedagogical value.

The study’s data told a clear story. Students with ASD in classrooms equipped with the NAO robot exhibited notably improved performance compared to students in regular classrooms. The researchers’ preliminary findings indicate that the robot significantly enhanced focus and classroom engagement among students with autism—two behavioral pillars on which nearly all other classroom learning depends. Improved attention, in turn, appears to translate into better educational performance and potentially enhanced social functioning. These outcomes align with a broader literature: studies of robot-mediated group instruction and robot-assisted psychosocial interventions have reported gains in attention, imitation, and social responsiveness, while systematic reviews of robotics protocols for students with autism have called for exactly this kind of ecologically valid classroom evidence.

The implications extend beyond special education. Classroom bonding is widely recognized as a foundation for healthy development, and children with autism are at elevated risk of disengagement from school environments that feel socially punishing. If a social robot can lower the barrier to participation—making group instruction feel approachable rather than threatening—the technology could help close an achievement gap that has proven stubbornly resistant to conventional approaches. The findings also carry practical weight for teachers: the robot-assisted model tested here positions NAO as an assistant that amplifies, rather than supplants, professional expertise. The teachers retained authority over pacing, content, and behavior management, while the robot contributed attention capture, novelty, and motivational energy that human instructors alone often struggle to sustain across a full group of autistic learners.

Cautions remain. This was a foundational study conducted at a single special education school, and the authors themselves describe it as a first step toward extended robot-assisted classroom sessions. Questions about long-term novelty effects—whether children’s fascination with a robot fades after weeks or months—persist in the literature, even as some long-term engagement studies suggest sustained benefit. Sample sizes in robot-assisted autism research are typically modest, and generalizing from one school to diverse educational systems will require replication. Ethical safeguards, however, were carefully observed: the study received approval from the Institutional Review Board of Suzhou University of Technology, written consent was obtained from guardians and cognitively capable children, and video data were encrypted after encoding to protect privacy.

Even with those caveats, the study marks a meaningful shift in the trajectory of social robotics for autism. For years, the field has demonstrated that robots can capture the attention of autistic children in controlled settings; the harder question has always been whether that magic survives contact with real classrooms, real curricula, and real group dynamics. By showing that a teacher-and-robot partnership can measurably improve focus and performance among students with autism in an actual school, the researchers provide the most convincing answer yet that social robots belong not just in therapy rooms, but at the front of the class. As extended sessions begin, the world may be watching a small humanoid robot help rewrite what inclusive, effective education looks like for children on the autism spectrum.

Subject of Research: The use of the NAO social robot to improve classroom performance and engagement of students with autism spectrum disorder

Article Title: Classroom Performance of Students with Autism in Interaction with the NAO Robot

Article References: Feng, H., Yang, Q., Lu, H., & Gong, S. (2026). Classroom Performance of Students with Autism in Interaction with the NAO Robot. Frontiers of Digital Education, 3(2), Article 9. https://doi.org/10.1007/s44366-026-0083-1

Image Credits: AI Generated

DOI: 10.1007/s44366-026-0083-1

Keywords: autism spectrum disorder, NAO robot, social robotics, robot-assisted classroom, special education, classroom engagement, human-robot interaction, educational technology, autism intervention, attention and focus, Classroom, Performance

The central insight behind the research is a concept the authors call Critical Decision Points, or CDPs. These are specific states in a shared task where two or more competing human strategies would prescribe clearly different next actions. Imagine a kitchen robot collaborating with a person on a recipe: if two different dishes the person might be preparing require identical actions up to a certain moment, watching the person gives the robot no new information. But if the robot guides the interaction so that the person reaches a junction where making soup demands reaching for a pot while making a salad demands reaching for a cutting board, the very next action unambiguously discloses the goal. CDPs are precisely those junctions, and they maximally reveal intent because the divergent strategies force a distinguishing choice.

To turn this intuition into a computable method, the researchers formalized CDPs using a goal-conditioned policy divergence measure. In plain terms, the team evaluated, for any given state of the world, how differently an optimal policy would behave depending on which goal the human is pursuing. High divergence means the state is highly informative: small differences in behavior at that state separate the candidate goals cleanly. Low divergence means the state tells the robot almost nothing, because all plausible goals call for the same action. By quantifying this divergence, the robot can rank future states by how much information they would generate about the human’s hidden objective.

Measuring informativeness, however, is only half the problem. The robot also has to accomplish the actual task, whether that means assembling ingredients in a kitchen or chasing down an opponent in a game. The team therefore embedded the CDP framework inside a Receding-Horizon Planner, a control architecture rooted in model predictive control that repeatedly looks ahead over a short window of future actions, optimizes a cost function, executes the first action, and then replans. The cost function balances two competing objectives: task progress, meaning the robot keeps moving the joint activity toward completion or toward winning, and information gain, meaning the robot steers toward states that maximize policy divergence and therefore reveal the human’s goal. The receding-horizon structure keeps the planning computationally tractable, because the robot never has to search the full branching tree of future interactions; it only reasons a few steps ahead at a time while still accounting for how its own actions will shape what the human does next.

What makes the work especially notable is that the same unified framework operates in two dramatically different social regimes. In the collaborative setting, the robot and human share a fully observable cooking task, and both parties benefit from smooth coordination. Here, steering the human toward CDPs serves partnership quality: the earlier the robot knows whether its partner is preparing one dish or another, the earlier it can hand over the right utensil, clear the correct counter space, or divide labor sensibly. Faster and more accurate goal inference translates directly into more fluent teamwork, reduced awkward pauses, and fewer mistaken anticipations of what the human wants next.

In the competitive setting, the tables turn entirely. The two agents play a partially observable hide-and-seek game in which the human’s goal is hidden from the robot, and revealing it works to the robot’s advantage rather than the human’s. Even here, the CDP-driven planner proved effective: by maneuvering the human into states where the seeker’s and hider’s strategies prescribe different moves, the robot extracted decisive evidence about the opponent’s hidden objective. The authors frame this in the tradition of opponent modeling, in which an agent builds and refines a predictive model of a rival’s intentions from observed behavior. Provoking opponents into informative situations has precedent in the game-playing literature, but this study is among the first to unify adversarial information gathering and cooperative assistance under a single planning objective.

Evaluation was deliberately thorough. The researchers tested their method in both domains in simulation and then transferred the approach to real robots interacting with actual human participants, a transition that notoriously breaks many planning algorithms due to perception noise, motion constraints, and unpredictable human behavior. Across both simulated and physical trials, the CDP-based planner inferred human goals more accurately and, crucially, earlier than baseline strategies that either pursued the task greedily or gathered information without shaping the human. The earlier a robot can commit to a confident estimate of a partner’s goal, the more useful actions it can take on that partner’s behalf, which is why the timing improvement matters as much as the accuracy improvement.

The implications stretch well beyond kitchens and games. Assistive robots in homes, factories, and hospitals constantly face the problem of deciphering what a person is trying to do, often from ambiguous and fragmentary evidence. A wheelchair-mounted manipulator that can gently reposition a shared workspace so that a user’s next movement disambiguates their intent could provide safer and more responsive support. Warehouse cobots could coordinate with human pickers more fluidly. Even autonomous vehicles, which increasingly reason about the intentions of human drivers, could benefit from planners that recognize which traffic situations are most informative about another driver’s route. The study builds on a growing body of research showing that influence is a two-way street: humans adapt to robots, and robots that account for this adaptation, rather than treating human behavior as fixed background noise, plan fundamentally better.

At the same time, the work raises questions the authors themselves situate within a broader research conversation about robots that influence humans. Steering a person’s choices, even mildly and for legitimate informative purposes, touches on autonomy and transparency, and the competitive results in particular show that the same mathematics that helps a helpful kitchen assistant can empower an adversarial agent. The researchers report no competing interests and ground their contribution firmly in the engineering goal of faster mutual understanding. As robots move from controlled labs into shared human spaces, frameworks like CDPs suggest a future in which machine teammates do not merely watch and guess, but intelligently arrange the world so that human intentions become legible — a capability that promises smoother collaboration, sharper competition, and a clearer science of how people and machines can read each other’s minds in real time.

The study also connects to a longer intellectual lineage in machine learning and control. The idea of learning a reward function or objective from observed behavior traces back to inverse reinforcement learning, which showed that an observer can recover what an agent is optimizing by watching its actions. The present work inverts that logic: instead of passively accumulating observations, the robot actively engineers the observation conditions, choosing its own actions so that the human’s behavior becomes maximally diagnostic. This places the method in the emerging field of influence-aware planning, where an agent treats its partner’s future decisions as part of the environment it can shape rather than as an exogenous process it must simply predict.

Methodologically, the reliance on a receding-horizon architecture reflects decades of practice in model predictive control, a technique originally developed in industrial process control and now widespread in robotics because it handles constraints and replans gracefully under uncertainty. Adapting this mature machinery to a social objective — trading off task progress against epistemic value — is a notable engineering contribution, since classical cost functions rarely encode anything like information gain about a partner’s intent.

The dual-domain evaluation also strengthens the claim of generality. Many goal-recognition systems are validated in a single task, leaving open whether the underlying measure of informativeness is task-specific. By demonstrating that the same policy-divergence criterion improves inference in a fully observable cooperative setting and a partially observable adversarial one, the authors provide evidence that Critical Decision Points are a domain-independent structural property of sequential decision problems, not an artifact of a particular scenario. The consistency of results across simulation and physical hardware further suggests the approach is robust to the perception noise and behavioral variability that typically degrade deployed interactive planners.

Subject of Research: Robot planning strategies that influence human behavior to accelerate goal inference during collaboration and competition.

Article Title: Robots influencing humans to reveal their goals during collaboration and competition

Article References: Ghose, D., Gitelson, O., Lewkowicz, M., Brawer, J., Roncone, A., Vazquez, M., & Scassellati, B. (2026). Robots influencing humans to reveal their goals during collaboration and competition. Autonomous Robots, 50(3), Article 38. https://doi.org/10.1007/s10514-026-10267-2

Image Credits: AI Generated

DOI: 10.1007/s10514-026-10267-2

Keywords: human-robot interaction, goal inference, critical decision points, receding horizon planning, social robotics, opponent modeling, human-robot collaboration, game theory, artificial intelligence, planning, shared autonomy, Robots

The findings come from a scoping review conducted by Rosabel Martinez-Roig, María Aragonés-González and Miguel Cazorla of the University of Alicante in Spain, published in the Journal of New Approaches in Educational Research. The team searched Scopus, Web of Science, ERIC-ProQuest and Dialnet for peer-reviewed empirical studies published between the robot’s commercial launch in 2014 and 1 March 2026, following the PRISMA-ScR reporting guidelines. Out of 1,448 initial records, and after duplicates were removed and 345 full texts assessed, only 13 studies met all the inclusion criteria. That narrow yield is itself telling: despite more than a decade of commercial availability and enthusiastic marketing, rigorously documented classroom deployments of Pepper remain rare, recent and geographically concentrated.

The technical appeal of the platform is easy to understand. Pepper is a social humanoid designed for human interaction, equipped with cameras, sensors, mobility, speech, gestures and expressive non-verbal cues such as head movements and eye contact. Unlike tablets or desktop tutoring software, social robots embody interaction; they can be perceived by children as social agents capable of reciprocity, which is precisely why researchers have speculated they might serve as tutors, learning companions or even pupils in ‘learning-by-teaching’ scenarios where students consolidate their own knowledge by explaining content to the machine. But the review found a persistent gap between that theoretical versatility and what actually happens in schools, where Pepper’s effective autonomy proved limited in most documented settings.

Geographically, the evidence is dominated by Europe, with nine of the thirteen studies (69.2 percent) conducted there, particularly in Italy, Germany, the Netherlands and Sweden. Germany was the most frequently represented country, with work such as Donnermann and colleagues’ long-term study of adaptive robotic tutors supporting university students with exam preparation and motivation. Four studies were carried out in the United States, including an ethnographic project by LeTendre and Gray that followed adolescents through a project-based learning unit, documenting how students themselves came to understand both the possibilities and the hard limits of human-robot interaction. The first qualifying empirical study did not appear until 2018, four years after the robot’s launch, underlining how youthful this research area remains.

The review reveals a striking imbalance in where and with whom the robot is deployed. Five studies focused on primary education, with children aged six to twelve, while four took place in higher education and two in early years settings; only one examined secondary education. Three studies centred on pupils with special educational needs, particularly autism spectrum disorder, extending a line of research suggesting that the predictable, patient interaction style of social robots can support communication and participation in inclusive contexts. Methodologically, however, the field is dominated by qualitative approaches: eight studies used case studies or interaction analysis, three used mixed methods and only two were predominantly quantitative experiments. Eight of the thirteen studies involved fewer than fifty participants, and not a single large-scale longitudinal study was identified.

Three pedagogical configurations dominated the corpus. In six studies, Pepper acted as a tutor or teaching assistant, delivering guidance, structured feedback or gamified quizzes, including quiz-based university learning in Germany. In four, the robot was cast as a learning partner or tutored agent, a role exemplified by Swedish research in which children taught mathematics to the robot and, in doing so, rehearsed and revised their own reasoning. In three studies, Pepper served as a social mediator in inclusive settings, supporting storytelling and second-language learning among migrant children and pupils with special educational needs. Curricular coverage was diverse but thin: mathematics appeared most often, followed by STEM and programming activities, language learning, social-emotional development in early years, and even a sustainability-focused serious game in which the robot improved Italian children’s attitudes toward recycling.

The benefits reported are consistent across the corpus, though the authors caution against over-interpreting them. Twelve of the thirteen studies — 92.3 percent — documented increased motivation and engagement when the robot was present, with pupils frequently showing more enjoyment than with tablet-based equivalents of the same tasks. Five studies recorded improvements in social skills or communicative participation, including greater verbal engagement in collaborative mathematics. Four reported measurable academic gains, from improved university grades in robot-assisted tutoring to more positive attitudes toward recycling. Three studies noted enhanced sustained attention during structured tasks. Yet the reviewers stress that novelty effects, small samples, short interventions and children’s prior expectations may all inflate these apparent benefits, and that none of the outcomes can be attributed to the robot’s presence alone.

The obstacles documented are equally consistent, and in many ways more illuminating. Technical failures were the most frequently reported limitation, mentioned in ten of thirteen studies: voice recognition struggles with children’s voices and ambient noise, software glitches interrupt activity sequences, and programming demands exceed what most schools can sustain. In more than half the studies, teachers had to provide constant technical supervision, stepping in to repair interactions when the robot misheard a pupil or froze mid-lesson. Curricular integration was weak, with most deployments framed as one-off activities, pilots or limited experiments rather than embedded components of standard teaching. Organisationally, cost, device scarcity and reliance on technical staff constrained uptake, and a small number of studies raised ethical questions about privacy, data collection and the risk of ‘robotised’ teaching, as well as classroom competition over limited robot time.

Perhaps the review’s central insight concerns the teacher. In seven studies, educators served as technical supervisors; in six, they acted as pedagogical mediators who structured activities, allocated interaction turns and regulated participation. Only two studies involved teachers directly in designing the intervention from the start, and just three mentioned any specific training in the pedagogical use of the robot — a gap the authors identify as one of the field’s most significant blind spots. Situations of ‘trouble and repair’, where adult intervention rescued collapsing child-robot interactions, recur throughout the literature, and earlier work in a school for autistic children explicitly framed Pepper as a teaching aid external to the classroom rather than an autonomous instructor. The distinction between the robot’s technical autonomy and its pedagogical autonomy, the authors argue, is critical: its capacity for action is always embedded in human, technical and ethical decisions.

The overall verdict is measured rather than sensational. Pepper’s embodiment, humanoid appearance and multimodal interaction can spark initial interest, but they do not by themselves guarantee meaningful learning; outcomes hinge on instructional design, student profile, technical stability and teacher mediation. The authors call for the field to move beyond technological novelty toward longitudinal studies that test whether early enthusiasm endures, deeper investigation of teacher training and design involvement, and direct comparisons of the robot against tablets, virtual agents and hands-on materials to establish where, if anywhere, it adds unique value. Until then, the evidence suggests that the social robot in the classroom is best understood not as a replacement for teachers but as a demanding, engaging and highly dependent teaching resource — one whose success is written, lesson by lesson, by the humans standing beside it.

The scoping review’s methodological choices help explain both its strengths and its boundaries. By restricting the analysis to a single robotic platform, the authors avoided a common problem in educational technology research, in which reviews lump together devices with very different degrees of physicality, mobility, expressiveness and autonomy, blurring the conditions under which each actually functions. Pepper’s specific configuration, combining a wheeled mobile base with a chest-mounted tablet, speech and gesture, makes it a useful case study in how multimodal embodiment is translated into classroom practice.

The timing of the corpus is also noteworthy. Because the first qualifying study appeared only in 2018, the entire body of empirical evidence postdates the robot’s commercial launch by several years, suggesting that early adoption in schools lagged well behind the platform’s availability. The concentration of research in primary and higher education, with secondary schooling nearly absent, may reflect practical factors such as timetable flexibility, curricular pressure and the willingness of younger children to engage with novel devices, though the review does not establish causal explanations for this distribution.

For practitioners, the most actionable finding may be the recurring pattern of ‘trouble and repair’. Since technical interruptions were near-universal, schools considering deployment should plan for adult supervision as a structural requirement rather than an occasional contingency. The scarcity of teacher involvement in intervention design, noted in only two studies, likewise points toward a simple improvement: co-designing robot activities with educators from the outset, and embedding them within existing curricular units instead of isolated demonstrations.

The authors’ call for comparative research deserves emphasis. Without head-to-head comparisons against tablets, virtual agents or conventional materials, the field cannot determine whether Pepper’s embodied presence produces effects that cheaper technologies cannot replicate, or whether the observed gains reflect novelty and expectation rather than any unique pedagogical property of the robot itself.

Subject of Research: A scoping review of the implementation of the Pepper social robot in formal educational contexts

Article Title: Exploring the implementation of the Pepper social robot in formal education: a scoping review

Article References: Martinez-Roig, R., Aragonés-González, M., & Cazorla, M. (2026). Exploring the implementation of the Pepper social robot in formal education: a scoping review. Journal of New Approaches in Educational Research, 15(1), Article 22. https://doi.org/10.1007/s44322-026-00072-1

Image Credits: AI Generated

DOI: 10.1007/s44322-026-00072-1

Keywords: social robotics, Pepper robot, formal education, scoping review, educational technology, human-robot interaction, inclusive education, teacher mediation, primary education, student motivation, autism spectrum disorder, STEM learning