During a recent visit to San Francisco, the author experienced a ride in a Waymo autonomous vehicle, appreciating the convenience and quiet solitude it offered. While the driverless car provided a comfortable, frictionless journey, it also highlighted a broader concern: technologies that remove human interaction can unintentionally diminish valuable social exchanges. This phenomenon, termed the 'Waymo effect,' describes how frictionless technology encourages opting out of human contact, often without recognizing what is lost.
In research, large language models (LLMs) function similarly to Waymo cars by offering instant, on-demand intellectual assistance without the complexities of human collaboration. Unlike human colleagues who bring diverse perspectives, challenge assumptions, and introduce unanticipated insights, LLMs respond only to direct prompts and do not independently question the researcher's approach. This absence of friction—the give-and-take of human interaction—is a core component of collaborative research.
Collaboration in research is costly and time-consuming, involving travel, scheduling, and negotiation. However, these frictions foster trust, serendipity, and intellectual diversity, which are crucial for innovation. Current incentive structures in academia, including funding cuts to collaborative activities, pressure for rapid publication, and the appeal of AI tools that require no credit sharing, make relying on AI more attractive. This dynamic risks a trend of 'decollaboration,' where researchers increasingly work in isolation aided by AI, potentially narrowing the diversity of ideas and slowing transformative breakthroughs.
Moreover, writing—a key part of thinking and refining ideas—is often expedited by AI, but this can short-circuit the deep cognitive processes that come from grappling with complex arguments and receiving critical feedback. Psychological research shows that certain difficulties in learning and thinking are beneficial, and removing these challenges may reduce the depth of understanding.
Experts emphasize that AI should augment rather than replace human researchers. The concept of 'pilot-in-command science' envisions researchers maintaining control and authority while using AI as a tool. To preserve the collaborative fabric of research, institutions and funders need to recognize and support the value of human interaction by investing in workshops, visits, and informal encounters that foster dialogue and diverse perspectives.
Ultimately, while AI offers undeniable benefits in accelerating research, the challenge lies in balancing efficiency with the preservation of collaboration. Without intentional efforts to fund and value human engagement, research risks becoming a smooth but isolated ride, where the critical role of human insight and challenge is quietly lost.