A new study asked 20 workers in hospitality, manufacturing, and care what they actually wanted from the robots they'd work alongside. Not what they feared. What they wanted.
The answers weren't about job security. They were about social fit.
Workers had clear, specific preferences: the robot should acknowledge when it had finished a task. It should signal when something had gone wrong. It should respond when spoken to. Not in a conversational, humanoid way. Just legibly. The way a new colleague in their first week would make eye contact before walking past your desk.
What they didn't want: the robot making decisions on its own.
That's the part worth pausing on. Not because it's surprising that workers want some control. Because of how precisely they wanted it calibrated. Not "give robots no autonomy." Not "robots should be supervised at all times." The nuance was contextual. In hospitality, a robot moving through a shared space should defer to the humans around it. In manufacturing, a robot performing a set task should complete it without interruption. In care settings, the stakes around independent action were highest of all.
They weren't applying one blanket rule. They were already doing the management calculation: what level of independence is appropriate here, in this role, in this moment?
That's a level of sophistication most robot deployment frameworks don't credit workers with having.
Here's what tends to happen instead. A new robot is introduced to a floor or a facility. There's a technical briefing. Safety protocols are reviewed. A vendor demo runs. And then the robot is just... present. It works. Mechanically, it performs. But it doesn't say hello. It doesn't signal completion in any way a human would recognise. It operates in the shared space as if everyone around it is an obstacle to navigate rather than a team to join.
Workers notice this. The study found they can articulate it precisely. What they're describing, when you map it against the research, is a failure of basic social skills. Not emotional intelligence. Not small talk. The signal that says: I'm aware you're here. I've finished. I need you to know something. The acknowledgment that makes shared space feel shared.
The deployment industry has been treating worker resistance to robots as a fear problem. Retrain for acceptance. Run change management. Show that the robot isn't taking anyone's job.
The actual problem might be a manners problem.
That's both more solvable and more embarrassing. Social legibility is a design choice. It can be built in. The fact that it often isn't suggests that most deployment decisions are made by people who don't spend their workday next to the robot. The workers who do have already worked out the social contract they'd accept.
Nobody has been asking them.
Source: Niessen et al. (2026). "Workers expect basic social skills but limited autonomy from future robots." Frontiers in Robotics and AI. Read the study →