Media literacy is changing – again. And this time the answers are not clear

At this point it feels like a truism that the majority of people who were involved on the critical side of edtech during the web 2.0 boom went on to do something interesting and unexpected.

Take Mike Caulfield. Back in the day his thing was the “federated wiki” – which added version control and forking to the read-write web shibboleth of the wiki, close enough to the open education sphere I orbited for us to have met up at a few conferences, shared a beer, and added each other to Google Reader (damn, I’m old).

These days he is better known for the SIFT method: a plausible, simple, and frankly enjoyable approach to thinking about the veracity of what you read online. If that’s a new one on you (it was even on the actual BBC a couple of years back) the outline version is as follows:

• S – Stop. Because the modern internet is powered by attention and lubricated by anxiety, the temptation is to react to what you see rather than think about. If something you have read or seen causes an emotional reaction, sharing or commenting based on a gut reaction is unlikely to help. So you pause.
• I – Investigate the source. Who created the thing that is being shared? Are they reputable (is it a media organisation you have heard of or one you think you have heard of)? Are they domain experts – is your climate horror coming from someone who usually shares game reviews and memes? If it is an organisation or business, what does it do (sales, advocacy) and who funds or inspires it? Mike’s key question here is “would you trust this source even if it says something you disliked or disagreed with?”.
• F – Find better coverage. If you are wavering about the source you have, find a better one. Look somewhere you do trust for more information on the story. Does anyone else even have it (exclusives aren’t really a web thing)?
• T – Trace the claim to its original context. If you are still on the fence, find out who did have it first? Do you trust them? Does it feel likely that an X account with eight numbers after the name would have the skinny on a massive new government policy?

It’s not perfect, but it represents a very healthy set of habits to get into. There’s a very good book (Verified) that is written for a general audience and does the idea a bit more justice. It came out in 2023. For people who teach this stuff (in research skills modules, or in your library) SIFT is pretty much the state of the art.

Trouble is, it was state of the art in 2023. Just like Substack has largely replaced Google Reader (easier to monetise), large language models are rapidly replacing social media and the traditional free-to-read web as a primary source for many people. Even the old gold standard of a Google Search now starts with an AI generated summary that will occasionally spout absolute nonsense, but with ChatGPT the most downloaded app on Android and AI technology seeping into the algorithms and digital assistants that define online activity.

SIFT is rapidly starting to look like the stuff I learned at university in the 1990s about dot org websites being more trustworthy than dot com websites, or the stuff I taught at university in the 2000s about being careful with wikipedia (would that our students used wikipedia now!) There’s no source, there’s no looking for other coverage, there’s no original – or if there is it is somewhere within a large gobbet of illegally obtained training material finessed via a boiler room in Khartoum filled with people trying to keep your favourite bot from declaring itself “mechaHitler.”

Last week the Foreign Affairs Committee – in a genuinely terrifying report on Disinformation Diplomacy that you should probably read – set out that a whole society approach to something called “media literacy” is a key component of our national security. Earlier this month we got a policy paper – A Safe, Informed, Digital Nation – from DSIT and DCMS: big on making links between strategies, short on defining what the interventions needed on media literacy actually are. For one of the biggest problems facing the UK (and, frankly, the rest of the world) it doesn’t feel like enough.

At best, media literacy here means something like the SIFT model: easy to grasp, straightforward to implement, and about three years too late. At worst, we poison the well even further: imagine factchecking an LLM-generated facebook post using an LLM designed to make you feel smart and clever for using an LLM. Or imagine a situation where people conclude that they can trust nothing so just go on what their more credulous mates say at the pub.

So, given that higher education is the kind of place where people are meant to learn to discriminate between the real world and Russian disinformation, how do we actually do that?

The problem with SIFT in this world is that the sharing source is very often an LLM, which obscures the original source. The kinds of media we still trust tend to be locked behind paywalls – and people don’t share links any more. As Caulfield puts it (on his – free to read! – substack):

The web isn’t dead, but much of the webbiness of it is gone. We live in a world of walled gardens, where the path to a website is almost never a link but instead a web search, and increasingly that web search is not bringing most people to sites either. That’s partially due to the structure of web search, but it’s also due to the fact there are very few free sources of information on the web anymore.

His draft proposals focus on the intelligent and skillful use of LLMs to do this work – and then verifying the sources and authorities it offers you when asked. It’s a few steps beyond “@grok – is this true?” but for me the underlying problems are the same. What is the LLM you are asking trained on, what are the biases and predispositions baked in to the answers you are getting?

Those very first models of an emerging web literacy were situated in a world of information scarcity – we asked set questions about what was in front of us (“does this sound plausible?”, “does it cite sources”?, “does it look reputable?”) because it was genuinely hard to find alternate sources. The abundance era – the one that we are coming to the end of – was the era of the alternate source: the old journalistic idea of double-sourcing only held for so long before it became clear that you could find at least two people putting forward even the clearest nonsense, and that the default early-mid web satirical voice was very often read as truth.

Post-abundance your media literacy tutor has to recalibrate again. The lazy newsgathering trope of the Twitter trend has become the Tik-Tok trend – harder to find and share, even more impenetrable when your aunt sees a short-form video of Angela Rayner eating puppies on Facebook. Links and sources are actively deprivileged on what remains of open social media – and the closed group (discord, the eternal WhatsApp chat, telegram) aims to keep you within those walls rather than off “doing your own research” beyond watching the YouTube short that has just been shared. And an LLM is always there to reassure you that you are, in fact, the messiah. And of course, the mainstream media (including Wonkhe: which I hear is both left-wing woke nonsense and a tool of university management) cannot be trusted.

This is a society-scale challenge – you’d look in vain in the recent policy suggestions for anything that begins to address it (and, honestly, the foreign propaganda is using the same approaches as anyone else who wants to share a harmful opinion online) beyond something that is likely to be a good few years behind SIFT. At best.

Universities should be in this space. University researchers and university libraries should be in this space. If you are, tell me. Because I’m worried.