Hallucinating responsibly: A perspective series on the future of neuroimaging research in the LLM era

An OSSIG initiative on the future of sustainable science

Mission

Large language models (LLMs) and other AI systems are already changing how science is done. This raises urgent questions about training, research integrity, open science, institutional responsibility and environmental sustainability. Many researchers are unsure whether these tools are overall beneficial or detrimental for science and scientists, or how to adapt to tools that are already part of their daily work. The neuroimaging community lacks a space to examine its own transformation critically and explicitly. This OSSIG initiative aims to provide it.

We don’t aim to advocate a particular position or prescribe best practices. Instead, we want to understand the ongoing changes, question the assumptions behind the uptake of LLMs in our day-to-day scientific activities, weigh opportunities and risks, and consider how researchers and institutions can actively shape what comes next. To do so, we will bring together people with different scientific, professional, institutional and philosophical perspectives, and make their reasoning and their potential disagreements visible.

Themes

  • Training and research practice: How are LLMs changing how researchers learn, build expertise, analyse data and produce knowledge? Which skills should trainees still develop themselves? How do LLMs change how you see your role as a researcher?
  • Open science and ownership of knowledge: What does openness mean when research outputs also train AI systems? How should we think about proprietary tools, open-weight models and control over scientific knowledge?
  • Sustainability: What are the environmental and computational costs of AI in research, and what would responsible adoption look like?
  • Institutional policy and the future of academia: What responsibilities come with adopting commercial AI systems? How can we protect scientific incentives, independence, accessibility and diversity of knowledge?

Format

The OSSIG editorial team will invite contributors at different career stages and with varied expertise to write a short perspective piece (~1,000 words) on the theme most relevant to them. Pieces will be published on the OHBM blog, with a selection of articles also published in The Transmitter in coordination with its editorial team.

Guidance for contributors: You may find it useful to address four levels:

  • Position: Where do you stand?
  • Reasoning: What evidence, assumptions or values support your position?
  • Implications: What should researchers, institutions or the scientific community consider doing differently?
  • Challenge: Which common assumption deserves reconsideration?

These are prompts, not a template. We especially welcome disagreement with mainstream views and unconventional perspectives. Where you disagree, please say at what level: facts and evidence, logic, values and priorities, or broader views of science and progress. Naming this helps people stop talking past one another.

Beyond the page

The publications are a starting point for a continuing conversation. Building on the series, we plan an annual facilitated discussion in the Open Science Room at OHBM. It will bring selected topics and contributors together in person and, importantly, invite the audience to respond, challenge arguments and share their own experiences. Our longer-term aim is a lasting space within the OHBM community to examine how AI is changing science, and what kind of scientific community we want to become.