Is this idea from AI or humans? A detector claims it can tell

Researchers have developed a detector that aims to tell the difference between ideas made by humans and those conjured by artificial intelligence.

Used with care, IdeaLens could help to gauge the extent of AI use in submissions to conferences and journals, which have become “overwhelmed” with artificially generated content, says Rishanth Rajendhran, a computer scientist at the University of Maryland in College Park and co-author of the preprint¹ describing IdeaLens.

Using AI to write scientific papers is becoming increasingly acceptable, but only “as long as the core idea” is the researcher’s own, he says.

Is it OK for AI to write science papers? Nature survey shows researchers are split

The detector works by analysing the ‘outline’ of a work rather than the text itself. First, it distils a text — for example, a scientific paper — into an outline, stripping it of style. Then it looks for patterns that are characteristic of either AI or human-generated ideas, such as in arguments and flow.

The detector was able to spot AI-derived ideas 91% of the time. It “surprised me that they can detect AI-written arguments” from just the outlines, says Nikhil Garg, a computer scientist at Cornell Tech in New York City, who calls the detector impressive.

Garg cautions that because the tool is fallible, it should not be used to judge the human authenticity of individual cases, and that the detector might become inaccurate when used on types of data that are different from those used in training. “An AI ‘tell’ in one domain may simply be good human writing in another,” he says.

Structural differences

The group’s earlier work found that human-written stories differ structurally from those written by AI and that different chatbots tend to use the same arguments when writing essays. Together this suggested that it might be possible to make a detector for AI ideas, says Rajendhran.