Cornell’s Yian Yin: Iterating After Rejection Beats Switching Ideas, Study Shows

Computational social Yian Yin of Cornell University utilized automated learning methods on both approved and denied grant applications to investigate how scientists learn from failure, discovering that researchers who continue their work and incorporate feedback achieve success at a faster rate than those who scrap their original ideas.

Applying Machine Learning to Research Failures

Yian Yin seeks to understand the dynamics of science itself, focusing particularly on how innovations happen and how they may be accelerated. “I want to understand what the secret sources or recipes of success are,” Yin, of Cornell University, told Science News. By examining both approved and rejected grant proposals with computer algorithms, he sought to figure out how scientists learn from failures.

What he found was that scientists fail, and they fail often.

Some researchers switched gears after facing multiple rejections. Others continued their research and incorporated feedback, finding success at a faster rate. Those who changed ideas or scrapped original proposals did not draw much from their experience.

How do scientists learn from failure? This researcher has clues
Photo: spherenexus.in

Teams That Iterate After Rejection Fail Faster Toward Success

The study revealed distinct patterns in how scientific teams handle adversity. “A group that’s not only failing better and better,” Yin said, as reported by SphereNexus, “is also failing faster and increasingly faster towards their eventual success.” This trajectory highlights that setbacks contain valuable structural data for teams willing to parse reviewer critiques rather than abandon hypotheses.

In contrast, investigators who completely abandoned their initial pathways gained little cumulative advantage from the peer-review rejection process. The data show that iterative refinement drives modern research ecosystems.

Generative Artificial Intelligence in Scientific Publishing

More recently, Yin has been investigating how generative AI is changing scientific publishing. An unprecedented influx of AI-generated manuscripts is now reaching academic periodicals, causing the evaluation system to struggle. More seemingly well-written papers that lack substance are getting through, he found.

Coming up with large-scale solutions requires looking at the issue in a measurable way, according to findings published by NCIJ Network. “Considering both the technological and social side of scientific ecosystems would be something that’s really beneficial for the ongoing public debate and discussion about how we accelerate science in the age of artificial intelligence,” Yin told Science News.

Cornell University, where Yin is based, was co-founded in 1865 by Ezra Cornell and Andrew Dickson White as New York’s land-grant institution. It remains one of only three private land-grant universities in the United States, operating state-supported contract colleges such as Agriculture and Life Sciences alongside its privately endowed divisions.

As of fall 2024, the university enrolled 16,128 undergraduate and 10,665 postgraduate students from 130 countries. Its affiliates include 63 Nobel laureates and 4 Turing Award winners, reflecting a long-standing emphasis on research that shapes fields from computer science to ecology.

How do scientists learn from failure? This researcher has clues
Photo: ncijnetwork.com

Translating Research Resilience Into Practice

  • Researchers who iterate on rejected proposals rather than discarding them reach success faster.
  • Generative AI tools are currently increasing the volume of superficial submissions, creating bottlenecks in peer review.
  • Analyzing scientific ecosystems requires measuring both social dynamics and technological impacts.
Response to Rejection Outcome on Research Trajectory
Continue and incorporate feedback Achieve success at a faster rate
Change ideas or scrap proposals Fail to leverage experience effectively

References

  • Science News reporting on Yian Yin’s computational social science research.
Photo of author

Health Desk

Health Desk at Archyde.

Tony Finau Ends 2026 Season With Missed Cut, Looks to 2027

Galaxy Z Fold 8: Wider Tablet Display, 4-App Multitasking & Hinge Kickstand