AI-generated content takes misinformation to a next level. Source criticism is so basic, deep and never lacks behind tools that are supposed to detect AI-generated content

Recently I had the pleasure to give a talk for 10 classes of Danish high school students. My talk was supposed to answer the question: How can we spot misinformation?

The somewhat pessimistic news is that it is increasingly difficult to spot misinformation in the age of AI-generated videos, images and audios. We as humans are used to believing what we can see and hear, and while manipulated footage and audio have always existed since the radio and television were invented, AI-generated content takes misinformation to a next level. It is hard to distinguish real content from fake content for the naked eye or our ears. For now, the trained eye can still spot AI-generated videos because of their occasional sloppiness, but that might very well disappear in near future, maybe even within a year or two. Where does that leave us as media consumers? How can we still trust what we see and what we hear?

Source Criticism does not Care about technology

I brought two ideas two the students. First and foremost, we have one powerful tool at our disposal: The good old source criticism toolbox that students usually learn about in their history lessons and Danish (in other countries English, German, French…) lessons. I argued that source criticism does not care about technology. The method operates at a level that is “deeper” than the layer of any given technology. Therefore, source criticism as a method never lacks behind the most recent technological developments. This distinguishes source criticism from technological tools that are supposed to detect AI-generated content. These technological tools are caught in an AI-arms race, always one step behind. You can use source criticism for everything: videos, AI-slop, newspaper articles, audio, images and so forth, despite of the fact that the method was developed more than a century ago, where books and newspaper were still the state of the art “information technology”.

After my short presentation the teacher sent the students into groups, where they were supposed to discuss my and other presenters’ talks. Then, the students shared their results with us, and we were supposed to discuss their input together on stage. Unfortunately, we never managed to discuss the one sentence that appeared again and again: “We do not believe in source criticism, because the method has not been developed for 100 years, and we need an update for modern technology”.

Sceptic students

That confused me a bit, because had I not just told them, that source criticism is so basic that it works for any technology?

Let’s have a look. Based what I learned about source criticism in school, you have to ask a number of questions when you look at the source at hand.

1. What kind of source are we looking at?
2. Who is the author?
3. When was the material created?
4. Is the source a primary or secondary source? (i.e. is our material the original or just a recital?)
5. Is the material a first-hand account of some event or a second-hand account?
6. Who is the intended audience?
7. What do other sources say?

You can always extend the analysis with more source-critical questions like “What is the core message?”, “What is the intended effect on the audience?” and so forth.

So, does it really work in the age of AI-generated content?

Yes, it does. A lot of AI-generated content on social media fails the 2. test. Who is the author? Oftentimes, it is impossible to find out, who is behind anonymous profiles on social media platforms. If you cannot determine the author, the source should instantly be dismissed as a potential source of truth.

Now one could argue that dissidents and whistleblowers often hide behind anonymous identities. That is a fair point. But the “average” person is usually not asked to judge the trustworthiness of whistleblowers and dissidents. Here, we rely on professional journalists and other experts, and we have done so already in the age of radio stations and print-newspapers. So: there is nothing new here.

In my view, the “Who is the author”-question easily debunks a large chunk of AI-generated profiles on social media today. It is impossible to figure out where the material is coming from, so we cannot trust it. In case of fake-profiles that seem to mimic real human beings, the “when” question becomes important. Was a profile created after the release of ChatGPT and friends? Time to be careful. What do other sources say? Well, that might be difficult in the case of AI-generated influencers, but in the case of war-reports or other fake news, it is still possible to resort to other sources and the more traditional media sources to fact-check information.

The Limits of Source Criticism

I am ready to admit, that it is not an easy task to spot fake content, and it is not getting easier. Even source criticism might reach its limit in a situation where the “What do other sources say”-question becomes nearly impossible to investigate for “normal” people, simply because we might end up in situations, where other AI-generated content is flooding platforms to a degree that gives users the impression that other sources are corroborating AI-generated content as facts.

In this case, the students might be right: Source criticism has reached its limit.

Luckily, I had a second recommendation: Go back to traditional media: Get yourself a newspaper and watch the news from a public service news station (if you can).

Of course, traditional media is not perfect either. Of course, we also need to be source critical in relation to traditional media. Of course, a lot of legitimate criticism can be launched at contemporary newspapers and public service stations, from clickbait on their websites to cuts on investigative journalism and experts. Of course, traditional media sometimes fall into the trap too and spread misinformation. And yes, FoxNews and other channels are more “traditional media” outlets, but unfortunately also to a large extent a misinformation machine.

But: Despite of all that, we must acknowledge that journalists are still experts in their field. They work professionally with news, research, fact-checking and a difficult media landscape every single day, and traditional media is regulated and has press ethics guidelines. As a consequence, journalists are also better equipped than most of us are to detect misinformation and to instead serve us properly researched news. Mistakes happen – but overall, quality newspapers and TV-stations are certainly a better choice than random SoMe channels.

Exceptions exist, and I am certainly not advocating for a completely uncritical stance toward traditional media.

All I am saying is that if we have difficulties to spot misinformation on social media platforms, and if we have difficulties of applying source criticism ourselves, we might be better off by leaving misinformation-prone channels, and by going back to good old newspapers and public service TV-stations.

If this happens at a larger scale, this might even have the positive side-effect that newspapers and TV-stations increase their income and are able to strengthen investigative journalism and expert journalism again, which is more important than ever before in the age of large-scale, AI-generated misinformation.

Photo: Vitaly Gariev, Unsplash