Hello,
It´s been a while…
If you have stayed subscribed through the silence, thank you. I really appreciate you still being here, and I hope you will continue to stay subscribed because I am planning new and better things here.
I’ve been thinking about the format of the newsletter and what I can realistically sustain. My previous approach was curating several AI findings into one compilation. I loved the variety, but it was hectic. So I’m changing the format. I’ll focus on one piece of research at a time, looking more closely at it and what the findings mean in real life.
And if you are reading the fAInding for the first time, welcome!
This is where I take complex AI research and turn it into accessible stories about what it actually means for people, organisations, and society. I introduce the research and the researchers behind it, explain what they found, and explore why it matters beyond the academic text.
TODAY
AI is helping scientists publish more. But is it narrowing science?
Artificial intelligence is becoming part of academic research. Scientists use AI to analyse data, search literature, write and edit text, generate code, identify patterns, and explore possible research directions.
The promise is clear
If researchers can work faster, perhaps scientific discovery can accelerate too. But a larger question follows: does AI help science explore more possibilities, or does it concentrate attention around the questions that are easiest for AI to handle?
A 2026 study in Nature examines this tension.
The paper, “Artificial intelligence tools expand scientists’ impact but contract science’s focus,” was written by Qianyue Hao, Fengli Xu, Yong Li, and James Evans.
The researchers analysed 41.3 million research papers across the natural sciences. They used a language model, validated against expert-labelled examples, to identify research that appeared to be AI-augmented.
Their main finding is a paradox
AI appears to expand the impact of individual scientists while narrowing the collective focus of science.
Scientists identified as engaging in AI-augmented research published 3.02 times as many papers, received 4.84 times as many citations, and became research project leaders 1.37 years earlier than those who did not fall into that category.
For an individual researcher, these are big advantages. AI may help scientists process information, analyse data, produce research outputs, and move more quickly through parts of the research process.
But when the researchers looked at science as a whole, they found a different pattern. AI adoption was associated with a 4.63% reduction in the collective volume of scientific topics studied and a 22% decrease in scientists’ engagement with one another.
The researchers suggest that scientists are moving toward areas with abundant data and clear benchmarks, that is, fields where AI can be applied and evaluated most easily. This may make research more efficient, but it could also mean that less measurable, data-poor, or unconventional questions receive less attention.
This is the difference between optimisation and exploration.
Optimisation improves an existing approach or solves a well-defined problem. Exploration means asking new questions, gathering unfamiliar data, and looking beyond established paths.
AI may be especially powerful at optimisation. The concern is that, if academic systems reward speed and measurable productivity, researchers may concentrate on the problems AI is already good at solving.
The study does not prove that AI alone caused these changes because it is a large-scale observational study based mainly on publication and citation patterns. Scientists who use AI may already differ from those who do not, for example, in their funding, institutions, fields, or access to data.
Still, the pattern raises an important question for universities, funders, and researchers: are we measuring more scientific output, or are we supporting a broader and more adventurous science?
The challenge is to design research environments in which AI also helps scientists explore neglected questions, create new datasets, and work across disciplinary boundaries.
The next time you hear that AI will accelerate scientific discovery, you might ask:
Accelerate discovery in which directions, and who decides what is worth discovering?
Source:
Hao, Q., Xu, F., Li, Y., and Evans, J. “Artificial intelligence tools expand scientists’ impact but contract science’s focus.” Nature, published 14 January 2026. Read the paper
Thank you for reading this new format.
Also, feel free to share any suggestions you have for improving this newsletter in any way.
QUESTIONS AND RECOMMENDATIONS
If you have questions about AI research, or recommendations for a finding, topic, or field I should explore next, I would love to hear from you.
Just reply to this email and let me know what you would like to see tackled.
If you are researching AI yourself and would like me to share and discuss your findings, you are very welcome to get in touch too.
Sharing is Caring
And if you know someone who might enjoy this kind of research translated into a shorter, more accessible format, please share the fAInding with them.
Am I an AI Positivist or Pessimist?
Last month, I presented my latest paper on algorithmic narrowing, about how embedding AI can quietly shrink what an organisation can notice, interpret, learn and act on.
In the end, I got some really interesting comments, questions, and things to think about. The one that stuck with me was when someone asked whether I'm an AI positivist or an AI pessimist.
I found that question really interesting. My findings leaned toward the more cautionary side of AI, but I still believe AI can do and be used for a lot of good. So I probably land somewhere in the middle. How about you?
▶ Was this email forwarded to you?
Keep an eye out over the coming weeks, too. I have some new things in the works, and I’m looking forward to sharing them with you.
Cheers,
Maryam