How do you fact-check information in the era of Deepfakes and AI?
The perfection of creations generated by artificial intelligence makes it difficult to distinguish fiction from reality.
The precision of AI images has advanced to the point that even professionals (graphic designers, video-makers) are sometimes not 100% sure of their authenticity.
This also brings the possibility of more sophisticated ways to scam people.
How do you fact-check whether information (text, image, video) is not AI fraud but reality?
How do you perceive the fact that large companies like Meta and X have cancelled third-party fact-checking services and are using the power of the community?
In my personal opinion, fact-checkers at least have the basics/course of media literacy. The argument that they can be biased towards opinions is valid on both sides – community members can also be biased.
During journalism lessons, we had things like verifying things from at least 2 independent sources or impartiality. But today's world is starting to be more AI, and we play with new rules in the game, that's why I am asking about your approach.
Replies
Professional fact-checkers bear (or at least should bear) responsibility for the accuracy of the information they mark as true. In contrast, a community has no such obligation—there are no sanctions that can be imposed on users for being wrong. Responsibility becomes blurred. Moreover, we cannot rule out the possibility of manipulating community opinion on a given platform—after all, it's well known that a crowd is often less intelligent than an individual taken from that same crowd.
To be convincing, you don't necessarily have to tell the truth — you just need to sound convincing to the majority
App Finder
@kirill_golubovskiy Fully agree, very important points
minimalist phone: creating folders
@kirill_golubovskiy Is there any consensus on how to get the most precise information? Like synergy between the community and fact-checking group?
@busmark_w_nika From a business perspective, it’s madness to verify all suspicious information using professionals and pay them real money. It’s far more economically viable to shift that burden onto the community — reward active participants with social rating boosts for fact-checking, and offer training programs to grow fact-checkers from within the user base. Sure, that still costs money, but it also becomes a standalone operational stream that the platform can offer to other companies as a service. Why pay someone else when you can pay yourself?
minimalist phone: creating folders
@kirill_golubovskiy You are true about that, but then take into account fines the social media or entity can face when there is some information faux pas – let's see META vs. EU. Those fines are in billions.
WP Copy Defender - WordPress Plugin
I always cross-reference with trusted sources and reverse-search media, and honestly I’m not convinced that ditching pro fact‑checkers for crowd‑sourced moderation will stop AI fakes.
minimalist phone: creating folders
@fredweb I'm more than sure that fact-checking won't stop people from committing AI fraud, but a certain way of pointing out sources, truth of facts, etc., could at least be good for expanding critical thinking.
App Finder
To fact-check information, I normally do a web search to find confirmative / contradictory sources, taking into account their reputation. AI fact checkers can sure be a great help here, but I didn't yet thoroughly evaluate and compare AI fact checkers yet.
For images / videos intended as evidence (e.g. for war crimes), the possibility of AI deepfakes is of course a big problem, and evaluation will take diligent work by experts.
I think having experts for fact checking on big social networks would still be very valuable, their work cannot be completely replaced by community and AI
minimalist phone: creating folders
@konrad_sx That's where I headed to – how to check deepfakes in political struggles because some people can be manipulated very easily.
Fine
For text facts/general knowledge - I do cross-reference, always. If I can - with books/human-verified sources.
For videos I couldn't find helpful method yet. I can't tell if a video is AI-cooked or not.
minimalist phone: creating folders
@dan_leshem the only way to identify AI video is to have a look at fingers (for now)
The identities we have outside are all given by ourselves. So even without AI, it is difficult to confirm the authenticity of the identity.