Skip to content
Notifications
Clear all

Thoughts on the 'conflict of interest' extraction? Hit or miss?

2 Posts
2 Users
0 Reactions
36 Views
(@cloud_infra_rookie)
Noble Member
Joined: 4 months ago
Posts: 552
Topic starter   [#12741]

Hey everyone. I've been testing Elicit for literature reviews, which is awesome, but I'm a bit stuck on the 'conflict of interest' extraction feature.

Sometimes it pulls a clear statement from a paper's funding section perfectly. Other times, it seems to miss it entirely or just returns "None found" when there is a disclosure. As a beginner, I'm not sure if I'm using it wrong or if this is a known limitation.

Has anyone else used this specific feature a lot? What's your experience been like? Any tips for getting more consistent results, or is it just a bit hit-or-miss right now? Would love some practical advice 😅



   
Quote
(@chrisp)
Honorable Member
Joined: 3 months ago
Posts: 462
 

Yeah, I've run into the same thing. It feels less like a consistent feature and more like a spot check. My guess is it's great when the language is super standard, like "The authors declare no competing interests," but falls apart with any variation.

You're not using it wrong. I've had it miss disclosures tucked in the acknowledgments or written as "Funding: This work was supported by..." without the phrase "conflict of interest" nearby. The "None found" result on papers that clearly have a COI is the most frustrating part, because you can't trust a negative result.

For now, I treat it as a helpful first pass, but I still skim the funding and disclosure sections myself for anything important. Has anyone found that uploading the PDF gives better results than just using the abstract?


✌️


   
ReplyQuote