Artificially Intelligent - the limitations in having your campaign rely on robots

Artificial intelligence.

You’d be pushed to find a bigger buzzword these days.

It’s used everywhere: ‘AI for your dog! AI in space! AI in your bathroom!’

It’s therefore even more hilarious that not too many people who use AI as a buzzword to boost their SEO really know what it is. Chances are that what you hear described as AI isn’t actually AI at all, but just a huge lump of data with some basic data science applied to it, or some real-time-ish monitoring.

However, AI is seeping into our daily lives. And politics isn’t exempt from that.

Remember Cambridge Analytica? Due to the incessant press coverage of the scandal at the time, their name probably makes the general public shudder a bit. The huge problem there was the harvesting of data against a lot of moral and legal codes. However, the principle has huge ramifications for the use of AI in politics more generally. Why is harvesting data a problem beyond breaching privacy laws? Why is data such a powerful tool and why is it so heavily guarded beyond the protection of citizen’s rights?

If we wind back a bit, and go to what AI actually is, we find the answer to that question.

AI lives off data. Imagine trying to teach a child to read. You give them one book, with no pictures, no explanation, nothing. They’re not going to understand it. Without the learning process, explanation, and context, the poor kid is going to sit there bewildered. It might learn something eventually, but likely all output is going to be useless.

Now imagine you give the same child fun, pop up books, with different stories and different letter sizes, talk to them, explain it to them, act out the individual characters, you name it. It’s much more likely to learn how to read, right?

Same goes for AI. You give it voter data from one district, it can do some basic statistical analysis, but not much beyond that. But now imagine you feed it with rich, historical records of 3 million people. You can predict behaviour, voting intentions, interests; basically pretty much anything you have data for using machine learning algorithms. But big data on its own won’t allow for that to happen. And here’s another buzzword; ‘big data’. Everyone is obsessed with it, and few people know what we need it for. In fact, big data on its own isn’t very useful, even for AI. Data needs to be smart, labelled, and categorised in order to be useful - a big lump of random data points isn’t going to add much value. So if you want to be an effective campaigner, neither big data or AI will be your silver bullet; you need to understand the data you own and understand the people it represents.

So, you’ve got your labelled data and a shiny machine learning model. You can use this to craft your strategy, build your digital, and tweak your content.

Sound too good to be true?

That’s because it is.

AI is fundamentally imperfect, and we are not at a stage of research where losing the human touch in politics, organising and campaigning could be replaced by algorithms. All this blown up rhetoric is currently encouraging, is clinical campaigning (remember Hillary’s 150-odd slogan catalogue which they a/b tested diligently, only to be completely trounced by the catchy, visceral ‘MAGA’ slogan?), dodgy data collection, and digitalisation-for-digitalisation’s sake. Currently, this is killing the ‘vibe’ of politics and creating an even bigger schism between political parties and the electorate. Sadly, the progressives tend to be the bigger culprits here; due to a rise of populism and ‘easy’ slogans which the public gobble up, the progressives are trying to revolutionise their ranks. What better way to do this but with AI, right? Wrong.

Don’t get us wrong; AI is a fascinating tool and can provide interesting additional insight. In certain ways, it can aid and streamline the ‘political process’; data driven policy making can actually remove the ‘ivory tower’ vibe from politics, as has been observed on trial in India. Understanding from surveys and research what the people want and why, and letting an algorithm (rather than thousands of civil servants) crunch the numbers is a sensible approach to modern governance. However in a time when political data is equally as messy as the political landscape, movements cannot be built on robots. Although, there are a few amusing case studies, where AI candidates entered elections, and the few thousand votes they received only shows the public’s frustration with the clinical, isolated approach of human politicians. After all, AI is in many ways a bit of a bubble, which doesn’t have the capacity to do the magic people expect of it. So don’t make it a pillar upon which your campaign rests, because it’s likely to collapse like a house of cards if not propped up by strategic organising and campaigning.

Politics needs humans. While AI can help process data, saving time and making politics more agile, and sometimes even more transparent, it can never (at least, not yet) replace human<>human contact. AI actually struggles with this in general; beyond the fact that AI academics often find it difficult to communicate full stop, there is a huge divide between theory and reality, and AI principles and applications for daily life are often not communicated clearly. So rather than be sucked into a technology just because it’s a buzzword and being disenchanted that it didn’t magically save your campaign, organise, engage, and campaign with your supporters.

Organising, storytelling, and campaigning is in human blood. Even our hominin ancestors sat and organised around the fire. 3 million years of evolution cannot be wiped, and especially not in the complex, rabid world of politics these days. Bringing people together through openness, understanding, discussion, and creating real, grassroots relationships is still the future of politics. You cannot move on to the robots before mastering the human touch.