Attitude data from 10,000 visitors, one mall, and a Spot quadruped

The RAI Institute ran a free public robot pop-up at the CambridgeSide mall in Boston across summer 2025, and the HRI paper that came out of it — published via IEEE Spectrum on 5 April 2026 — is more useful to construction and BIM professionals than the headline might suggest. The short finding: a few minutes of hands-on control of a Boston Dynamics Spot shifted comfort scores upward across every deployment context tested, consistently and at statistical significance after correcting for multiple comparisons.

Roughly 10,000 visitors came through. Ten percent drove Spot through a custom obstacle arena and filled in pre/post surveys. More than 65% had already seen Spot online — but most had never been in the same room as one. That gap between mediated familiarity and lived experience is exactly what the study was designed to measure.

←TODAY: Spot has been commercially available since 2020; ANYmal by ANYrobotics — an ETH Zürich spin-off — is actively deployed on European construction sites for autonomous inspection scanning.
→3012: Social acceptance, not sensor resolution, is the rate-limiting variable on robot integration into occupied buildings.
Fulcrum: The attitude shift happens in the hands, not in the brochure — which means deployment strategy is a design problem.

The system behind the signal

The survey targeted two dimensions: comfort (how at ease would you feel encountering this robot in a given setting?) and suitability (how well would it function there?). Five contexts — factory, home, hospital, office, outdoor/disaster — were chosen specifically to span the range from high public acceptance (industrial, emergency response) to well-documented ambivalence (domestic, healthcare).

The largest attitude shift appeared in the outdoor/disaster context. Participants already rated Spot as suitable for search-and-rescue; they just weren’t comfortable with it. That discomfort tracks directly to media portrayals of quadruped robots in military settings — the Boston Dynamics viral videos that circulated 2018–2022 did real and measurable damage to public comfort in non-industrial scenarios. A few minutes of joystick control partially reversed that. The factory context, which already had the highest baseline comfort of all five, showed no statistically significant gain — a sensible ceiling effect.

The controller design is worth noting: custom-built on an adaptive video game controller, large-button layout, accessible from age 2 to over 90. This is universal design applied to HRI methodology, not just UX polish. It means the attitude data spans a genuine demographic range, not a self-selected tech-comfortable cohort.

On the working desk this week

For architects and project managers specifying site robotics — ANYmal for construction inspection, Spot for as-built scanning feeding into BIM point-cloud pipelines — the RAI Institute study surfaces a variable that rarely appears in procurement conversations: worker and occupant acceptance as a deployment bottleneck. A robot that passes every technical benchmark and fails the site team’s comfort threshold will be parked within a month. The research team, led by Dawn Wendell and Hae Won Park (whose parallel work at MIT’s Personal Robots Group focuses on social robot companions for aging populations), is essentially arguing that experiential introduction is a corrective to media-shaped fear. That argument has a direct construction-site analogue.

ANYrobotics, as an ETH Zürich spin-off now active across European industrial inspection, sits squarely in the DACH professional orbit. Their inclusion in the Boston pop-up alongside RAI Institute hardware is not incidental — it signals that the HRI research community is treating these platforms as the ones that will actually reach occupied buildings, not lab prototypes.

The trade-off worth stating plainly: this study used a controlled, gamified 10-minute session in a shopping mall. Sustained comfort in an actual hospital corridor or domestic kitchen — where the robot is not under the user’s direct control — is a different and harder problem. The paper does not claim otherwise, but vendors deploying these findings to justify rapid rollout should be pressed on that gap.

Atelier: If your office is preparing a Wettbewerb or feasibility study that includes robotic site inspection or occupant-facing robotics, the RAI Institute methodology offers a ready-made template: scenario-specific comfort/suitability split, pre/post interaction measurement, and controlled environment theming. Running a small internal version with your client’s facility management team before procurement could surface resistance early — and cheaply.

Read the IEEE Spectrum piece by Wendell, Park, and Leone, then look up ANYrobotics’ published deployment case studies for European construction sites. Bring both to your next project kick-off that includes any autonomous scanning workflow. The question to put on the table: who on this site will actually operate the robot, and have they touched one yet?

Source: IEEE Spectrum