One-Week Smartphone Intervention for Autistic Adults Falls Short in First Trial

A brief smartphone-based intervention designed to sharpen social attention in autistic adults retained every participant who started it, yet failed to meet benchmarks for adherence and acceptability and produced no detectable improvements in social-emotional functioning, according to a randomized trial published in the Journal of Autism and Developmental Disorders. The study, led by China I. Parenteau, Allison S. Nahmias, and Matthew D. Lerner, is the first to test an ecological momentary intervention (EMI) targeting social attention in autistic adults, and its mixed results offer a sobering counterpoint to the enthusiasm that has surrounded low-intensity, mobile-delivered supports.

Ecological momentary interventions build on ecological momentary assessment, a method in which participants receive mobile prompts in their natural environments and report on recent behaviors, thoughts, or experiences, sometimes at randomized times of day. Where assessment stops at measurement, an EMI embeds brief therapeutic content—reminders, emotion-focused prompts, or behavioral suggestions—directly into daily life. In non-autistic populations, EMIs have shown strong feasibility and acceptability across domains including depression, post-traumatic stress, and substance use. Because autistic adults often face limited access to scalable social-emotional supports, particularly as services drop off during the transition to adulthood, the researchers argued that a brief, flexible, phone-delivered approach was well suited for feasibility testing in this population.

The intervention itself was deliberately minimal. Thirty-one autistic adults, aged 18 to 45 with an average age of about 26, were randomly assigned to one of three conditions. For seven to eight days, all participants used a smartphone-based Google Form to report their social interactions, defined as back-and-forth exchanges lasting longer than five seconds in which each person’s behavior directly responded to the others. Participants in the two active conditions answered a follow-up question after each reported interaction: those in the facial expression condition logged their conversational partner’s initial facial expression, while those in the hair color condition logged the partner’s hair color, a comparison designed to separate social-emotional attention from general social attention. A passive control group answered only the basic interaction question. Everyone also received twelve daily text message reminders—two at fixed times and ten at semi-random intervals—prompting them to submit reports.

The rationale rested on the idea that social attention, the ability to orient toward socially relevant cues, is a foundational process supporting downstream social-emotional skills such as emotion recognition and empathy. Facial expressions in particular are processed by specialized brain regions, and electrophysiological research has shown that an early face-processing component of the brain’s electrical activity, the N170, is often delayed in autism. Prior intervention research with autistic youth had shown that social-emotional functioning, including emotion recognition and even N170 response speed, can improve through intensive programs, giving the team reason to think that repeated, real-world prompting might strengthen engagement with social cues in adults.

On retention, the results were unambiguously positive. All 31 participants completed the study, a 100 percent rate that significantly exceeded the 20 percent maximum dropout benchmark the researchers set in advance. The authors attribute this to the intervention’s brevity, the short daily survey—at most two questions per entry, compared with one to 38 questions in prior ecological momentary studies with autistic individuals—and the mobile, flexible delivery format. But the picture darkened on adherence. Participants self-reported logging only about 61 percent of interactions under signal-contingent reporting, in which entries followed automated reminders, and about 54 percent under event-contingent reporting, in which participants logged interactions on their own initiative. Both figures fell significantly below the a priori 80 percent adequacy benchmark.

Acceptability also missed its threshold. The team computed a net acceptability score for each participant by counting aspects of the study the participant liked and subtracting the aspects they disliked, drawn from open-ended responses. Sixty-eight percent of participants reported scores of zero or above, but this did not differ significantly from the 50 percent benchmark in a binomial test, so the a priori acceptability criterion was not met. Qualitative feedback was genuinely mixed: some participants praised increased self-awareness, easy reporting, and helpful reminders, while others found the logging burdensome, questioned its purpose, or complained of too many reminders. Multiple participants reported technical difficulties, including inconsistent or missing reminders and trouble accessing the reporting form. Notably, the open-ended prompts often elicited comments about the study broadly—staff interactions, the EEG session, the commute—rather than the intervention itself, and the authors caution that a negativity bias in such responses may have underrepresented positive experiences.

Equivalence testing showed that utilization, adherence, and acceptability were statistically indistinguishable across the three conditions, though the authors urge caution given the small sample and wide, power-derived equivalence bounds. The efficacy results were equally null. Regression models adjusting for baseline scores found no significant effects of condition on behavioral emotion recognition measured by the DANVA-2, on self-report questionnaires covering empathy, mutuality in close relationships, and trait emotional intelligence, or on neural outcomes. One uncorrected effect did emerge: post-intervention N170 latency at electrode site P7 was faster in the combined facial expression and hair color conditions relative to passive control, a potentially intriguing signal given that N170 latency is often delayed in autistic populations. However, the effect did not survive correction for multiple comparisons, and the authors are explicit that no detectable changes in social-emotional functioning were produced.

The authors identify several methodological constraints that likely shaped these outcomes. The mixed event- and signal-contingent design made true utilization impossible to calculate, since there is no objective way to know how many social interactions a participant actually experienced; retrospective self-reported adherence, vulnerable to recall bias, served as the closest proxy. Prior autism diagnoses were self-reported and not verified against clinical records, and seven participants—about 23 percent of the sample—were included on the basis of an elevated score on the Autism Spectrum Quotient alone, a screening tool rather than a diagnostic instrument, though sensitivity analyses excluding them left the overall pattern unchanged. The small sample reduced statistical power, and the one-week duration, chosen to minimize burden during feasibility testing, limited the study’s capacity to detect meaningful change. The authors stress that efficacy findings from this window should be read as preliminary rather than confirmatory.

Looking forward, the team proposes concrete refinements: clearer benchmarks for expected logging frequency, fuller explanations of why each prompt matters, compensation tied to logging, visual supports within the logging interface, and more reliable reminder delivery. They also suggest extending the intervention toward 10 to 14 days, the point at which engagement in prior momentary intervention research tends to decline, and increasing intensity, for example by prompting participants to track a partner’s expressions across the beginning, middle, and end of an interaction or to connect facial expressions with verbal cues. Because several participants described increased self-awareness as a benefit, future iterations may add self-awareness as an outcome. The authors also recommend partnering with autistic adults to redesign the intervention, given that neither adherence nor acceptability thresholds were achieved.

The study’s broader message is one of tempered realism about low-intensity digital supports. A one-week protocol with minimal active ingredients achieved remarkable retention, suggesting autistic adults will stay engaged with brief, mobile, low-burden research. But retention alone did not translate into adherence, acceptability, or measurable change in social-emotional functioning, whether measured behaviorally, through self-report, or in the brain’s face-processing response. As the authors conclude, features of this intervention may need substantial refinement to align with autistic adults’ preferences, and the potential of ecologically momentary approaches for this population remains, for now, an open question rather than a demonstrated benefit.

Subject of Research: Feasibility, acceptability, and preliminary efficacy of a one-week ecological momentary intervention targeting social attention in autistic adults

Article Title: Feasibility, Acceptability, and Efficacy of an Ecological Momentary Intervention for Autistic Adults

Article References: Parenteau, C. I., Nahmias, A. S., & Lerner, M. D. (2026). Feasibility, Acceptability, and Efficacy of an Ecological Momentary Intervention for Autistic Adults. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07574-z

Image Credits: AI Generated

DOI: 10.1007/s10803-026-07574-z

Keywords: autism, autistic adults, ecological momentary intervention, mobile health, social attention, social-emotional functioning, adherence, acceptability, randomized trial, N170, EEG, digital intervention