How AI is Already Winning Elections
The technology is changing GOTV campaigns
You’ve got to get the vote out. That’s what really wins elections. Sure, the media campaigns do matter, as well as the credibility of any candidate, but that counts for nothing if your supporters stay at home.
GOTV techniques can increase turnout by anything between four and nine per cent (study here), depending on the tactics you employ and the country you’re canvassing in.
The stakes are big, White House or Westminster big, and for decades political parties have been refining their data and software to produce more accurate results for door-knocks, leaflet drops and phone calls.
As it stands, the story of the US election campaign is yet to be fully told. But with Trump’s ground campaign being outsourced to Elon Musk’s America PAC (outlined more extensively here), you can reasonably expect that AI technology had some role in the Republican victory.
In the UK, meanwhile, it’s starting to become clearer on how Labour used Large Language Models to secure a landslide in July.
As previously reported in this newsletter, and yet to be picked-up elsewhere, Keir Starmer’s party was experimenting with new AI-supported social media techniques, which allowed them to interact with pro-Labour voters if they left a positive message on the leader’s or party’s platforms. The idea was to boost the party’s GOTV.
Tim Ross and Rachel Wearmouth, authors of Landslide: The Inside Story of the 2024 Election, have also uncovered another piece of technological innovation: Labour’s so-called ‘feedback loop’.
The AI-driven GOTV system was apparently championed by Hollie Ridley, who oversaw Labour’s ground game. According to Ross and Wearmouth, here’s how the the ‘feedback loop’ worked:
“Party organisers on the ground would collect and submit comments from voters, stories from regional or local news outlets, and results from surveys.
“An AI tool would then collect this information into a written report. These reports would be poured over by campaign chiefs the next morning to help inform the orders [Morgan] McSweeney and his deputies would send back into the field that day.”
Labour canvassers were also encouraged to adopt a techie approach to door-knocking, following the ‘Persuasion Pathways’ script to get the most engagement out of their conversations.
A training ad from 2023 boasted about ‘Persuasion Pathways’ promised to teach volunteers “innovative ways” of canvassing through a “new and exciting” method.
In reality, that meant opening a conversation with a voter so they could share their frustrations, concerns and top issues first before being asked for a one-out-of-ten ranking on if they plan to vote Labour or not.
Due to their flexibility, the systems worked well in the now promiscuous environment of British electoral politics – a phenomenon which has seen voters shed old party loyalties and back insurgents on the right and left.
The Conservatives, who were forced to defend traditional safe seats which had become super-marginals, were found wanting. Some Tory MPs, as Ross and Wearmouth point out, just couldn’t find their voters due to years of inaction and complacency.
And GOTV tech can only do so much. If you lack volunteers, have poor party morale and are struggling to compete in the top three issues (typically, the economy, the NHS and immigration in the UK), you’re not going to win an election.
By comparison and reflecting on some earlier political innovations, Vote Leave built its own multi-input voter intention software in the run-up to the 2016 EU Referendum.
Here’s how the ‘Voter Intention Collection System’ (VICS) was used according to Dominic Cummings (link to full article):
“VL had targeted GOTV activity in 96% of constituencies and marginal-seat style GOTV operations in half all constituencies.
“Activists knocked on about half a million doors (low estimate) before GOTV. On 23 June they knocked on about 1.5 million doors and probably interacted with about half a million people…
“By the end of May activists were collecting and entering around 7 of 16 25,000 voting intentions per week, which later increased in the first two weeks of June to around 35,000 VIs per week.
“This is a rate of ~140,000 per month so if we could have done that from January we could have got ~106 canvassed VIs to add to those from direct mail and the data science models (the latter of which provided most of the doors to be knocked for GOTV).
“Data from canvassing helped improve the accuracy of the data science models which in turn helped improve the efficiency of canvassing and all of it helped target social media too.”
The advent of generative AI is only going to ramp-up the competitiveness of the GOTV software. But it’s only as good as any party that uses it.
What next for The Observer?
If The Observer isn’t sold to Tortoise Media, what future will it have as part of the Guardian Media Group, which clearly wants to offload the title?
The outlet’s now ex-editor Paul Webster recently took to Twitter to argue that The Observer made a “net contribution” to the parent company, only later to say this:
“Even once added expenses are included, as they have been in the company's recent shameful attempt to bundle the paper out of the door as quickly as possible, and account is taken of shared resources we use, the losses are miniscule set against the company's £1.3 billion cash pile.”
This clearly doesn’t make sense. As for Tortoise, its accounts show that it made a loss of £4.6 million in 2022, bringing in revenues of £6.2 million. It sounds like a great business, with a fair amount of costs.
But the James Harding-run operation is now briefing that it will make a profit in the final quarter of this year.
The former editor of The Times and director of BBC News has promised to invest £25 million in the combined entity over the next five years, £5 million of which is reportedly expected to come from future profits.
Based on other reports (Sunday Times / FT), here’s what the plan looks like:
Launch a digital paywall for The Observer
Bulk-up the weekly’s sports and business coverage
Combine the readership of Tortoise and The Observer
Break-even by 2027
Secure 179,000 total subs by 2029 and become profitable
Time for social graph removers
Since there is now a plethora of social (and non-social) platforms, it’s time for someone to step-up and create a social graph removals company — sort of like those blokes with big vans who help you move house.
For me, that’s the main thing that’s stopping me from joining Bluesky. After more than a decade on Twitter, doing it all again looks like an arduous and unthankful task.
Since Substack is now positioning itself towards a more social media-esque platform with the roll-out of ‘followers’, conversation-loving internet users have to pump-out content across different websites. Oh, how I miss the days of MSN Messenger.
🎙️ The Political Press Box
Find my latest long-form audio interviews here.
📧 Contact
For high-praise, tips or gripes, please contact the editor at iansilvera@gmail.com or via @ianjsilvera. Follow on LinkedIn here.
207 can be found here
206 can be found here
205 can be found here
204 can be found here
203 can be found here
202 can be found here