Grammarly is using our identities without permission
Grammarly’s “expert review” feature offers to give users writing advice “inspired by” subject experts, including recently deceased professors, like Wired reported Wednesday. When I tried this feature myself, I found some experts who were a surprise for a different reason: one of them was my boss.
The AI-generated comments included comments that appeared to come from The edgeEditor-in-Chief Nilay Patel, along with Editor-in-Chief David Pierce and Editors-in-Chief Sean Hollister and Tom Warren, neither of whom gave Grammarly permission to include them in the “expert reviews.”
The feature, launched in August, claims to help you “refine your message through the lens of industry-relevant perspectives.” When users select the “expert review” button in the Grammarly sidebar, it analyzes their writing and surfaces AI-generated suggestions “inspired by” relevant experts. These “industry-relevant perspectives” include, among others, Stephen King, Neil deGrasse Tyson, and Carl Sagan.
The edge found many other tech journalists named in the article, including former Edge editors Casey Newton and Joanna Stern, former Edge the writer Monica Chin, WiredThis is Lauren Goode, Bloombergit’s Mark Gurman and Jason Schreier, THE The New York Times Kashmir Hill, The AtlanticIt’s Kaitlyn Tiffany, PC gamerIt’s Wes Fenlon, GizmodoIt’s Raymond Wong, Digital Foundry founder Richard Leadbetter, Tom’s guide editor-in-chief Mark Spoonauer, former Rock Paper Shotgun Katharine Castle, editor and former IGN news director Kat Bailey. Some experts’ descriptions contain inaccuracies, such as outdated job titles, which could have been accurately updated if Superhuman had asked these individuals for permission to reference their work.
In a statement to The edgeAlex Gay, VP of Product and Corporate Marketing at Superhuman, parent company of Grammarly, commented: “The Expert Review Agent does not claim endorsement or direct participation by these experts; it provides suggestions inspired by the work of experts and directs users to influential voices that they can then explore in more depth. »
When asked if Superhuman plans to notify the people named in its AI feature or ask their permission, Gay responded: “Expert Review experts appear because their published work is publicly available and widely cited. »
However, the work of the experts proved difficult to “deepen”. The feature crashed frequently and its “sources” were linked to spammy copies of legitimate websites or other archived copies that are not the actual source page.
“Expert Review experts appear because their published work is publicly available and widely cited.”
Some sources even accessed completely unrelated links that were not written by the person they were supposedly an example of, potentially indicating that suggestions Grammarly’s AI offers with one person’s name may be based on another person’s work. This is only apparent if users click “see more” to expand the suggestions, then click the “source” button at the end of the suggestion.
Additionally, the way the suggestions are presented could be misleading. In Google Docs, suggestions resemble comments from real users, seemingly simulating the experience of receiving edits from the AI-imitated expert. A Grammarly AI suggestion “inspired by” Edge Editor Sean Hollister wanted to add a parenthetical with context already included elsewhere. The only problem is that I was actually edited by the real Sean Hollister, who prefers to avoid repetitive or unnecessary explanations while using simple wording and organization.
If I had taken that advice and followed it up with him, the real Sean probably would have removed the parenthetical Grammarly suggested. An AI may be able to ingest large quantities of someone’s writing and learn to imitate it, sure, but the same strategy can’t teach an AI how to modify the way that person would, based solely on the writing they’ve published, even if you give the robot a checkmark logo and call it an “expert.”
