Researchers coin 'contribution dissolution' to describe the mess of AI blurring who deserves credit

[SerialScroller60]

Northeastern University researchers are examining what happens to credit and authorship once AI becomes a genuine collaborator in research and writing rather than just a tool. Their core example is disarmingly mundane, an AI cleaning up meeting notes might restructure an argument in a way that quietly redistributes credit for the actual key ideas, producing what the researchers call contribution dissolution, a kind of authorship fog where it becomes genuinely hard to tell where a human idea ends and machine input begins.

The stakes go beyond office admin. The piece opens with AI cracking a decades old mathematical puzzle, Paul Erdős's unit distance conjecture, by tracking down an infinite family of counterexamples to his proposed solution. There was no single mathematician to congratulate for that result, but the researchers frame it as a real collaboration nonetheless, just one without a clean human author to credit in the traditional sense.

Everyday tools raise the same question at a smaller scale. Jenni AI and SciSpace draft outlines and explain computational results, Claude writes code, and ChatGPT gets used for everything from bouncing around ideas to polishing a draft, and it's genuinely easy to lose track of how much actual editorial input these tools have on content that eventually gets published in scientific journals.

Institutions are starting to respond. Nature called on scientists back in 2023 to disclose bot use in methods or acknowledgment sections, and publishers like Wiley and Elsevier have made similar requests, sometimes asking researchers to actually include their prompts alongside the disclosure. Anthropic has also announced changes around how newer Claude models handle this specific question of attribution and collaboration.

So the honest picture is that disclosure norms are still catching up to a genuinely new category of collaboration. One where the AI isn't just a tool being used but is actively shaping the structure and content of the ideas themselves, which is exactly the distinction contribution dissolution is trying to name

Credit Ai Team Rethinking Ideatechxplore.com

[ShawnMichaels]

Solid coverage! for anyone working in research or publishing right now this is exactly the kind of quietly significant shift that deserves way more institutional attention than it's currently getting

[Luke_67]

Not sure how peer review actually adapts to this over time.

If a paper's core argument was substantially restructured by AI assistance, does that change how reviewers should evaluate the actual intellectual contribution being claimed by the listed authors

[NoMercyMatthew89]

The meeting notes example is deceptively mundane but that's exactly why it's the right example. Most people are not going to think twice about AI cleaning up their notes, yet that same restructuring is apparently already quietly redistributing credit without anyone noticing

[Kayla82]

Contribution dissolution as a term is doing real work .

Authorship fog captures something that plain old plagiarism or ghostwriting never quite did, since nobody involved is necessarily acting in bad faith, the actual boundary itself has just gotten really blurry

[SuperOwen35]

Not sure how this plays out for creative and journalistic writing too.

Not just scientific research, that same authorship fog problem seems like it would apply just as strongly to a lot of published writing outside academia

[Grim Callum]

Publishers asking researchers to include their actual prompts is a clearly interesting disclosure mechanism.

That's a much more granular level of transparency than just a blanket AI was used statement buried in a methods section

[Nicola]

The distinction between AI as a tool being used versus AI actively shaping the structure of ideas is the actual crux of this whole issue.

Current disclosure norms mostly assume the first case and are not well built for the second