The Waymo effect: how AI is quietly making research less collaborative

Your research process may become more isolated, and the novelty of your findings could decrease due to AI tools.

The "Waymo effect," named after self-driving car technology, describes how technologies that eliminate human friction are perceived as pure gain because the costs are visible and the benefits are hidden. The author, Daniel Hook, argues that this effect is now diminishing collaboration in research. Similar to how Waymo cars offer a frictionless ride, large language models (LLMs) provide researchers with instant, tireless assistance. This convenience, however, comes at the cost of the unpredictable challenges and serendipitous discoveries that arise from human collaboration. LLMs, unlike human collaborators, do not introduce their own agendas, challenge assumptions unprompted, or offer views from outside a researcher's immediate bubble. The research ecosystem, with its emphasis on speed and output, inadvertently incentivizes this "decollaboration," making the frictionless LLM a rational choice despite the long-term impact on the community and the novelty of research.

The "Waymo effect" highlights how technologies that remove human interaction, like Waymo's driverless cars, are embraced for their convenience, masking the loss of subtle human benefits. This principle is now being applied to research through large language models (LLMs), which offer researchers instant availability and assistance, eliminating the friction of human collaboration. The author posits that this trend, termed "decollaboration," prioritizes individual productivity and speed over the unpredictable, yet vital, challenges and serendipity that human collaborators provide. The research community's incentive structures, focused on output and velocity, further encourage this shift toward AI-driven research isolation.

This "decollaboration" has significant implications for the future of research, potentially leading to a narrowing of ideas and a decrease in truly novel work. The research ecosystem's current reward system, which emphasizes speed and measurable output, favors the efficiency of LLMs, inadvertently penalizing the slower, more complex process of genuine collaboration. As researchers increasingly rely on LLMs, the social fabric of research—maintained through argument, apprenticeship, and informal discussions—risks unraveling. This shift represents a global experiment trading serendipity for throughput, with potentially profound consequences for scientific advancement and the diversity of thought.