Political ads on Twitter: good, bad or indifferent?
Plus the trouble with mushrooms (or at least the AI-generated guides to them)
Twitter1 is going to allow political advertising again. In one respect, this is hardly surprising: the company is absolutely on its knees, financially speaking, and political advertising probably pays better than hole-in-the-wall Alibaba dropshipping vendors trying to sell things that aren’t legal in the country where they’re shown. And of course that also fits into Elon Musk’s insistence that his social media plaything should be a place where anyone can say anything, as long as it’s not about his whereabouts.
The blogpost about the reintroduction (“Supporting peoples’ right to accurate and safe political discourse on X”) insists of course that it’s all going to be done really, really well—like everything else on modern Twitter:
Building on our commitment to free expression, we are also going to allow political advertising. Starting in the U.S., we’ll continue to apply specific policies to paid-for promoted political posts. This will include prohibiting the promotion of false or misleading content, including false or misleading information intended to undermine public confidence in an election, while seeking to preserve free and open political discourse. We’ll also provide a global advertising transparency center so that everyone can review political posts being promoted on X, in addition to robust screening processes to ensure only eligible groups and campaigns are able to advertise.
“Prohibiting the promotion of false or misleading content”? Pretty much every political ad is false or misleading in some way, unless all it’s telling you is the candidate’s name and their party, so it’s going to be quite the show watching politicians complaining about having ads thrown back. (Somehow I expect that right-wing American politicians will be fine, but those further left will find objections raised.)
Twitter banned political advertising globally in October 2019, with the explanation given by then-CEO Jack Dorsey being that Social Media Is So Powerful:
“While internet advertising is incredibly powerful and very effective for commercial advertisers, that power brings significant risks to politics, where it can be used to influence votes to affect the lives of millions,” Dorsey wrote in a multi-tweet thread explaining the decision. “Political message reach should be earned, not bought.”
The contrast is with Facebook, which has pretty much always allowed political advertising, and—unlike what New Twitter claims it will do—doesn’t do fact-checking, nor ban political ads, unless they try to mislead people about where or when you can vote.
This led to an entertaining exchange in Congress, when Mark Zuckerberg was cross-examined by Alexandria Ocasio-Cortez, as I recounted in Social Warming:
Ocasio-Cortez: ‘Would I be able to run advertisements targeting Republicans in primaries saying they voted for the Green New Deal [an expensive scheme to tackle climate change]? If you’re not fact-checking political advertisements – I’m trying to understand the bounds here of what’s fair game.’
Zuckerberg wasn’t sure: ‘I think, probably?’
AOC: ‘Do you see a potential problem here, with a complete lack of fact-checking on political advertisements?’
MZ: ‘Well, Congresswoman, I think lying is bad, and I think if you were to run an ad that had a lie that would be bad, that’s different from it being, from in our position it being the right thing to do to prevent your constituents from seeing that you have lied—’
AOC: ‘So you will take down lies, or you won’t take down lies?’
MZ: ‘I believe in a democracy I think people should be able to see for themselves what politicians that they may or may not vote for say, so they judge their character for themselves.’
AOC: ‘You may flag that it’s wrong, but you won’t take it down.’
MZ: ‘It depends on the context . . .’ Zuckerberg trailed off.
It was at almost exactly this time—probably the same Congressional inquiry—that Dorsey made his announcement about Twitter.
What this doesn’t answer, though, is the crucial question: how effective is political advertising on social media? Dorsey, in his tweet thread, didn’t cite any research.
What I found from researching the book is that political ads on social media (hell, anywhere) are not very effective. But that doesn’t mean they’re ineffective. The presidential elections in the Philippines in 2016 were peppered with misinformation revolving around Facebook posts; the Brexit referendum saw the biggest ever voter turnout, driven by people who had never voted in their life but had been energised by propaganda on (again) Facebook; and the US 2016 presidential election is famous for having been the forum for microtargeting of ads aimed at people in Wisconsin, Michigan and Pennsylvania, three states which turned out to be crucial to Donald Trump winning the electoral college, and hence the election.
The trouble is, getting people interested enough in politics that they actually take notice of what politicians are saying—even at election times—is remarkably hard. You pay attention to all the politicians being interviewed, which is why you grow weary of them trotting out the same line again and again. But the saying in politics is that by the time you’re absolutely sick of repeating that line, in TV studios, on the stump, to news journalists, to radio journalists, to passersby, to your long-suffering spouse—by that time, the public is just starting to pick up that you’ve got a slogan.
The same is generally true for political advertising, on social networks as anywhere else. The key difference, though, is that those ads can be picked up by the networks’ users and amplified—positively or negatively—and then gain a life of their own. (Facebook prices adverts partly based on a calculation of their potential virality: the more likely to go viral, the cheaper.)
We’ve seen examples of this: in April a UK Labour Party advert, which it then tweeted (neat way around any ad ban), suggested that the Conservative Party leader (and PM) Rishi Sunak didn’t want adults convicted of sexually assaulting children to go to prison. Inevitably, it got a Community Note (I like them), though it’s pretty equivocal:
The resulting storm around the claim roiled social media for quite a while—all of which had the effect of showing the ad to more people. Even with the Community Note, you can bet lots of people just absorbed the message. And it got a lot of action. 22.3 million views (if Twitter’s metrics can be trusted, though they can’t really), 800+ retweets, and the Likes nearly keeping pace with the 5,000-odd quote tweets (which I think we can assume are probably negative). This is the crucial difference: if you hit the social network’s G-spot, its denizens will do all the work of spreading your message.
But does that change people’s minds, or make them more determined to vote—the two things that politicians want from any campaign? There’s not a lot of research. A team from the Oxford Internet Institute published a paper in 2019, looking at the effect in the 2015 and 2017 UK elections:
We find that Twitter-based campaigning does seem to help win votes. The impact of Twitter use is small, though comparable with campaign spending. Our data suggest that social media campaign effects are achieved through using Twitter as a broadcast mechanism. Despite much literature encouraging politicians to engage with social platforms in an interactive fashion, we find no evidence that this style of communication improves electoral outcomes.
There’s also a nice bit of evidence for social warming from misleading content, in a January 2022 paper titled “Misleading political advertising fuels incivility online: a social network analysis of 2020 U.S. presidential election campaign video comments on YouTube”. (Take note, Twitter and Facebook!)
The most recent study I can find was published in January this year, published in Nature, titled “A 2 million-person, campaign-wide field experiment shows how digital advertising affects voter turnout”. The abstract says:
Treatment group participants were exposed to an 8-month-long advertising programme delivered via social media, designed to persuade people to vote against Donald Trump and for Joe Biden. We found no evidence that the programme increased or decreased turnout on average.
There was some really minimal differential effect, but nothing to write home about. The BPS website has a longer writeup, but the finding—for those who spent all that money—is pretty grim:
[the saturation campaign] increased voting among people who had been identified as leaning towards Biden by only 0.4 percentage points, while it decreased voting among Trump-leaners by just 0.3 percentage points. This effect was “much smaller than pundits and media commentators often assume”, the team writes.
So maybe we shouldn’t fret about Twitter allowing political advertising back. All that seems likely to happen is that it will narrow the wide gap between the company’s outgoings and revenue. And, of course, provide us with lots of annoying political ads to complain about.
Glimpses of the AI tsunami
(Of the what? Read here. And then the update.)
• If you go down to the woods today, don’t take one of the books popping up on Amazon which claim to help you choose between deadly and safe mushrooms: they’re AI-generated and untrustworthy. The linked story does a good job of pointing out how you can tell that a book on this topic is probably chatbottery. So maybe we need a book to help us pick the book to help us pick mushrooms. But soon, there would be untrustworthy AI-generated books pretending to be books to help us pick the book to help us pick mushrooms. So we’d need a book to.. actually, shall we make a meal without mushrooms?
• Google is introducing a “watermark” that it says will identify an image that has been AI-generated—but only if the creator uses Google’s Vertex AI system. Demis Hassabis, of DeepMind, says the intention is to expand it further, and that it resists cropping or resizing, while not affecting the appearance of the photo. To me, that’s steganography—hiding a message in a picture—and probably defeated by the same processes that defeat it such as file conversion. We will see.
• AI for me but not for thee: JP Morgan Chase, the bank, tells people that ChatGPT et al are a “paradigm shift” and “transformative”. However: “JPMorgan has blocked access to ChatGPT from its computers and told its 300,000 workers not to put any bank information into the chatbot or other generative AI tools.”
• You can buy Social Warming in paperback, hardback or ebook via One World Publications, or order it through your friendly local bookstore. Or listen to me read it on Audible.
You could also sign up for The Overspill, a daily list of links with short extracts and brief commentary on things I find interesting in tech, science, medicine, politics and any other topic that takes my fancy.
• I’m the proposed Class Representative for a lawsuit against Google in the UK on behalf of publishers. If you sold open display ads in the UK after 2014, you might be a member of the class. Read more at Googleadclaim.co.uk. (Or see the press release.)
• Back next week! Or leave a comment here, or in the Substack chat, or Substack Notes, or write it in a letter and put it in a bottle so that The Police write a song about it after it falls through a wormhole and goes back in time.
I’m not calling it “X”; if that offends you, maybe there’s a browser extension or something which will do the replacement for you seamlessly.
The strongest argument to me for the overall effectiveness of political advertising, is that the opposite is to claim all the professional politicians don't know their jobs. That they're all raising and then wasting massive amount of money on nothing. Isn't this exactly the sort of situation where everyone is supposed to derisively shout "Dunning-Kruger!" at anyone contradicting the real experts? (i.e., those people who have actually won elections). This actually seems to be a classic situation of that type - those who have the strongest empirical results and the most stake in an outcome have a very strong consensus on a model, are told they are wrong by a bunch of pundits and think-tankers using very poor theoretical research. In a different social context, I suspect this arrangement would lead to much mockery by the chattering class.
Now, the above is not absolute proof (practioners can be wrong). But I find it much worth pondering.