Computer scientist Ben Shneiderman argues against the move towards full robotic automation, saying such moves absolve humans of ethical responsibilities
A computer scientist argues that the quest for fully automated robots is misguided, perhaps even dangerous. His decades of warnings are gaining more attention. Tweets: @cigitalgem , @edyson , @ctricot , @nytimesbusiness , and @jhagel Tweets: Gary McGraw / @cigitalgem : Food for thought as we rush headlong towards AI https://www.nytimes.com/... Esther Dyson / @edyson : As @benbendc says, metaphors matter. Perhaps the ideal “autonomous” computer is more like a horse than a machine, with a mind of its own but in constant communication with its owner. A Case for Cooperation Between Machines and Humans https://www.nytimes.com/... https://twitter.com/... Christophe Tricot / @ctricot : A Case for Cooperation Between Machines and Humans The quest for fully automated robots is misguided, perhaps even dangerous. Even dangerous. Robots should collaborate with humans rather than replace them. #ai #Humans https://www.nytimes.com/... https://twitter.com/... @nytimesbusiness : Should machines ever be truly allowed to make bid decisions on their own? For years, some researchers have said, “No.” That argument is getting louder. https://www.nytimes.com/... John Hagel / @jhagel : Should we be striving for autonomous machines or collaborative machines? The debate continues as reported by @markoff https://www.nytimes.com/...
Context & Ripple Effects
Ben Shneiderman's warning lands against a backdrop the Times has been documenting for years: executives at Davos describing a competitive race toward automation regardless of what it does to workers, and the Pentagon-adjacent debate over autonomous weapons that would make lethal decisions without a human. His argument — that full autonomy doesn't just displace labor but dissolves moral responsibility — gives both debates a single framing.
The pickup matters because of who is amplifying it: Esther Dyson's 'more like a horse than a machine' reframing and security researcher Gary McGraw sharing the piece show the critique moving from HCI circles into the mainstream AI conversation, just as later coverage shows researchers divided on AI risk while global regulatory frameworks lag behind.
First-order effects
- Executives currently justifying automation as a competitive necessity now face a named counter-position from a credible computer scientist: that designing humans out also designs accountability out.
- Dyson's horse metaphor offers product designers a concrete alternative vocabulary — supervised partnership rather than 'autonomous' — that directly challenges how these systems are marketed.
Second-order effects
- The autonomous-weapons argument over imbuing robots with lethal decision-making extends beyond defense: any vendor selling full autonomy must now answer whether its system can carry blame, not just perform tasks.
- If buyers start treating human oversight as an ethical requirement rather than an efficiency cost, automation roadmaps built on removing people — like those described at Davos — face pricing and procurement pressure to keep humans in the loop.
Third-order effects
- With regulators still catching up on AI risk broadly, Shneiderman-style arguments give policymakers a template for assigning liability: systems designed for human-machine cooperation leave a responsible party, while fully autonomous designs leave a gap.
- The pattern points toward a structural fork in AI development — between an autonomy-maximizing camp and a collaboration-first camp — with the same fault line visible later in disputes over whether researchers should claim their systems have minds at all.
The trend: As automation accelerates ahead of regulation, the defining design fight in AI is shifting from capability to where responsibility sits — full autonomy versus enforced human oversight.