Hands-On Robot Control Shifts Public Comfort More Than Passive Exposure
A mall pop-up where 1,000 people drove a Spot robot for a few minutes produced statistically significant comfort gains across every social context tested — including the ones robots are most likely to enter next: homes, hospitals, and offices.
Explanation
The RAI Institute (the research arm spun out of Boston Dynamics' founding team) set up a free robotics exhibit in a Cambridge, MA mall in summer 2025. The centerpiece was "Drive-a-Spot" — a small obstacle course where anyone could pilot Boston Dynamics' Spot quadruped using an oversized adaptive controller accessible to ages 2 through 90+. About 10,000 people visited; 10% drove the robot and filled out before/after surveys.
The headline result: comfort scores rose across all five tested contexts (factory, home, hospital, office, outdoor/disaster) after just a short hands-on session. The gains were small-to-moderate in size but statistically robust. The biggest jump came in outdoor/disaster scenarios — people already thought Spot would be useful there, they just didn't feel good about it, likely due to military-robot imagery in media. A few minutes at the controls partially dissolved that.
More interesting than the comfort numbers: perceived suitability gains were largest in home, office, and hospital — exactly the environments where skepticism was highest at baseline. And the effect generalized: people who drove through a home-themed arena also rated hospitals and offices as more robot-appropriate. That suggests hands-on control changes something deeper than context-specific familiarity — it updates a person's mental model of what robots can actually do.
Post-session emotions skewed hard positive: 74% reported excitement, only 12% nervousness. Notably, after driving, the share of people who wanted robots for "entertainment and play" jumped from 7.5% to 19.4%, while references to hazardous labor declined. People stopped imagining Spot as a factory tool and started imagining it as a companion.
The practical implication is blunt: if the robotics industry wants public acceptance in domestic and healthcare spaces, passive content — videos, articles, exhibits — may be the wrong lever. Letting people actually operate the machine works faster and more broadly.
The study, presented at HRI 2026 in Edinburgh, is one of the larger in-the-wild human-robot interaction datasets collected outside a lab setting. ~1,000 participants completing paired pre/post surveys across a demographically broad, self-selected mall sample is not a controlled RCT, but it's meaningfully more ecologically valid than the typical 20-person university study.
The within-subjects design (same participant rates comfort before and after driving) controls for individual baseline differences, and the team corrected for multiple comparisons across five contexts — a methodological step many HRI papers skip. Effect sizes are described as "small to moderate," which is honest; this isn't a conversion event, it's an incremental shift.
The generalization finding is the most theoretically interesting result. Suitability gains in home, hospital, and office contexts weren't confined to participants who drove through those themed arenas — they appeared across arena conditions. This is consistent with schema-update models of attitude change: direct agency over a system revises the underlying capability model, not just the context-specific association. Prior HRI work (e.g., Heerink et al. on UTAUT in social robots) has documented familiarity effects, but typically via repeated exposure over days, not a single short session.
The demographic data adds nuance without resolving the equity question. Men started with higher baseline comfort than women across all contexts; interaction narrowed the gap only in factory and office settings. Children showed stronger office comfort gains but persistent factory skepticism — plausibly a domain-familiarity artifact rather than a robot-specific effect. Prior Spot drivers (mostly robotics professionals) started higher but were caught up by novices post-session, which argues against a ceiling effect and for the intervention's broad applicability.
Key open questions the paper doesn't answer: (1) Durability — do attitude shifts persist at 1 week, 1 month? (2) Selection bias — mall visitors who choose to drive a robot are not the general public; the most robot-averse people likely walked past. (3) Conflict of interest — this is RAI Institute studying its own exhibit with its own robot. The paper is peer-reviewed (HRI 2026), which mitigates but doesn't eliminate that concern. (4) Behavioral validity — comfort ratings are not behavioral intentions, and neither predicts actual adoption. Watch for follow-up longitudinal work; that's where this line of research either earns its claims or doesn't.
Reality meter
Why this score?
Trust Layer A single short hands-on session controlling a Spot robot raises public comfort and perceived suitability across all tested deployment contexts, including the socially sensitive ones where robots are most likely to be deployed next.
A single short hands-on session controlling a Spot robot raises public comfort and perceived suitability across all tested deployment contexts, including the socially sensitive ones where robots are most likely to be deployed next.
- ~10,000 visitors attended the mall pop-up; ~1,000 drove Spot and completed paired pre/post surveys.
- Comfort scores increased significantly across all five contexts (factory, home, hospital, office, outdoor/disaster) after the driving session, with effects described as small to moderate but statistically robust after multiple-comparison correction.
- The largest comfort gain was in outdoor/disaster scenarios, attributed to media-driven military-robot associations being partially dissolved by direct control.
- Suitability gains were largest in home, office, and hospital — the highest-skepticism contexts — and generalized across arena themes, suggesting a capability-model update rather than context-specific familiarity.
- Post-session, the share of participants wanting robots for entertainment/play rose from 7.5% to 19.4%; 74% reported excitement and only 12% nervousness.
- Participants with no prior Spot exposure caught up to experienced drivers (robotics professionals) after the session.
- Self-selection bias: mall visitors who opt into driving a robot are likely already more curious or positively disposed than the general public; the most averse individuals are not captured.
- Conflict of interest: the study was designed, run, and reported by RAI Institute using its own exhibit and its own robot; peer review at HRI 2026 mitigates but does not eliminate this.
- No durability data: the paper measures attitude shifts immediately post-session; whether gains persist over days or weeks is entirely unknown.
The study is peer-reviewed (HRI 2026), uses a within-subjects design with multiple-comparison correction, and reports effect sizes honestly as small-to-moderate — methodological transparency that supports a moderate-to-high reality score.
The source is written by the researchers themselves and frames the pop-up as a clear success; the self-selection bias and absence of longitudinal data are acknowledged but not foregrounded, warranting a moderate hype flag.
If the generalization and durability of these attitude shifts hold up in follow-on work, the implications for robotics deployment strategy — prioritizing hands-on public access over media campaigns — are concrete and near-term, justifying a meaningful impact score.
- 1 source on file
- Avg trust 40/100
- Trust 40/100
Time horizon
Community read
Glossary
- within-subjects design
- A research method where the same participants are measured multiple times under different conditions (in this case, before and after driving the robot), allowing researchers to control for individual differences by comparing each person to themselves.
- ecologically valid
- Research findings that reflect real-world conditions and behaviors outside of controlled laboratory settings, making the results more applicable to actual human experiences.
- schema-update models
- Psychological theories explaining how direct experience with something can revise a person's underlying mental model or beliefs about that thing, rather than just changing their feelings in that specific situation.
- ceiling effect
- A limitation in research where participants' scores are already so high at the start that there is little room for improvement, making it difficult to measure the true impact of an intervention.
- selection bias
- A systematic error that occurs when the participants studied are not representative of the broader population, such as when only people willing to volunteer are included in a study.
- behavioral validity
- The degree to which self-reported measures (like comfort ratings) actually predict or correspond to real-world actions and decisions people make.
What's your read?
Your read shapes future topic weighting.
Your vote feeds topic weights, community direction and future prioritisation. Open community direction
Sources
Optional Submit a prediction Optional: add your prediction on the core question if you like.
Prediction
Will a follow-up study confirm that comfort gains from hands-on robot interaction persist for at least one month after the experience?