Preprint models how LLMs can reshape scientific incentives toward faster, lower-quality work
_akpiper · x · 2026-07-22
The paper argues LLMs can change research incentives, not just productivity
A new preprint by Carl Bergstrom, Kevin Gross, and MJ Crockett models how LLMs affect the way researchers allocate time and effort.
- The core claim is counterintuitive: instead of simply freeing scientists to do better work, LLMs may push the system toward more output, faster iteration, and lower quality.
- The figure models optimal policies across discovery and development phases, showing how opportunity cost and project value determine whether a project gets developed or abandoned.
- One panel suggests that if an LLM speeds up the discretionary development phase, projects that do get developed may be pursued more thoroughly.
- But other panels show that reducing the time needed for discovery or other stages can raise the opportunity cost of time, changing which projects are worth continuing and how much effort gets invested.
In short, the paper frames LLMs as a force that can reshape scientific incentives at the system level, not merely a tool that boosts individual productivity.
More from AGI Musings
- Podcast maps the open-model race across Kimi, Qwen, GLM and Chinese labs — natolambert · 2026-07-22
- The New Yorker links the AI gender gap to the parenting gender gap — newyorker · 2026-07-22
- Alignment trade-off: doing good and obeying users may not both maximize — ctjlewis · 2026-07-22
- Africa AI Workshop: Building Agentic AI Without Frontier Model Dependency — ChinasaTOkolo · 2026-07-22
- Self-evolving Lean proof agents reach 45.1% on miniF2F with coevolving benchmarks — omarsar0 · 2026-07-22
- AI benchmark-maxing ignores speed, cost and compute power — DevToD4 · 2026-07-22