The Waymo Effect: How Frontier Autonomous Systems Are Quietly Making Scientific Research Less Collaborative

In an influential essay published in Research Agenda, Daniel Hook, Chief Scientific Officer at the Holtzbrinck Group, formulated a psychological diagnosis for modern academia: the 'Waymo Effect.' Named after the serene comfort of stepping into an autonomous robotaxi in San Francisco, the term captures what occurs when technology removes the friction of dealing with another human being, and we experience that absence as pure gain—because the costs of friction were always visible, while its subtle benefits were not.

In intellectual life, large language models are rapidly becoming the driverless capsules of scientific discovery. A human collaborator is messy: delayed by teaching schedules, burdened by author-order negotiations, and prone to arguing that your foundational premise is flawed. An AI assistant, by contrast, is on-call at 2 AM, demands zero author credit, challenges arguments only as far as prompted, and delivers clean prose on demand.

Under systemic publish-or-perish pressures, defunded travel budgets, and institutional metrics that worship publication velocity, turning to the machine has become a rational choice for researchers. Yet Hook warns this frictionless convenience masks an existential crisis: 'decollaboration.'

Science is not a sterile production line converting compute into citations; it is a social fabric maintained through passionate argument and corridor serendipity. Team-science literature has long demonstrated that while large teams incrementally develop established paradigms, atypical, small pairings of researchers across disparate disciplines drive the most disruptive breakthroughs.

When researchers swap human friction for compliant AI models, individual output climbs while the diversity of ideas contracts. Because models optimize for probabilistic consensus, independent researchers are quietly nudged along similar cognitive corridors. The scientific enterprise is effectively trading serendipity for throughput—maximizing paper volume while narrowing the variance of collective discovery.

The deeper cost strikes at the cognitive heart of research: the act of writing. Drawing on Robert Bjork's concept of 'desirable difficulties,' psychologists know that the effortful struggle to articulate a concept in one's own words is not an obstacle to understanding, but the mechanism that creates it. Writing is where logical gaps reveal themselves.

Outsourcing drafting to LLMs does not accelerate thinking; it bypasses it. A research ecosystem can appear thriving by every visible metric—higher publication rates, faster turnarounds, spotless syntax—while the underlying capacity for original critique quietly atrophies below critical thresholds.

Echoing Dashun Wang's framing in Nature, the imperative is not to abandon AI, but to reject 'passenger-in-comfort' science in favor of 'pilot-in-command' stewardship. Researchers must remain captains who actively design dissent into agentic workflows rather than passively accepting machine consensus.

Ultimately, university leaders and grant funding bodies must realign institutional incentives: they must actively 'fund the friction.' This requires investing in physical workshops, visiting fellowships, and unscripted time where serendipitous human disagreement can flourish. In an age where automated output is cheap, the human friction of thinking together has become science's most precious commodity.

• The Waymo effect: how AI is quietly making research less collaborative — Research Agenda News
• The Waymo effect: how AI is quietly making research less collaborative — Wonkhe
• A Framework for Pilot-in-Command AI Science — Nature