Did algorithms help Trump and Brexit to win? This question is on everyone’s lips after the issue was surfaced by Swiss journalists. The news basically claimed that Trump’s campaign used highly sophisticated algorithms, offered by a UK-based company Cambridge Analytica, to influence the voting behaviour of millions of Americans, based on data from their Facebook profiles and other sources.
Image courtesy of Stuart Miles at FreeDigitalPhotos.net
Soon after, Scout painted a very gloomy picture of the large-scale “Weaponized AI Propaganda Machine”, involving microtargeting, bots and fake news. Buzzfeed’s take on the subject was more skeptical, especially regarding the actual role of the algorithms in the election outcome.
Without access to the data from the companies and organisations involved, it’s impossible to know what kind of data and algorithms were actually used, and what their impact was. But I think that this is not the most important question anyway.
Instead, I would ask whether the claimed scenario is technically possible, and what we should do about it.
Personal data enables personalised targeting
Data scientists in the advertising industry have been developing data-driven personalised targeting tools for years now. They have become masters in both collecting personal data from social media and other sources, and applying advanced machine learning and artificial intelligence systems to influence consumer choices. And the algorithms1 keep getting better, learning from constant feedback.
Now it is not hard to imagine someone with enough money and understanding to apply the same approach to advance political interests.
The core issue here that I want to emphasise is the access to massive loads of personal data that are essential for efforts to influence people on large scale. A lot of data is legally available, especially in the US, and much more is available for those not playing by the rules. As Professor Jonathan Rust (director of Cambridge University’s Psychometric Centre) put it recently in Guardian:
“The danger of not having regulation around the sort of data you can get from Facebook and elsewhere is clear. With this, a computer can actually do psychology, it can predict and potentially control human behaviour. It’s what the scientologists try to do but much more powerful. It’s how you brainwash someone. It’s incredibly dangerous.
“It’s no exaggeration to say that minds can be changed. Behaviour can be predicted and controlled. I find it incredibly scary. I really do. Because nobody has really followed through on the possible consequences of all this. People don’t know it’s happening to them. Their attitudes are being changed behind their backs.”
In other words, AI systems that feed on personal data have made it possible to influence people’s behaviour.
It’s important to note that the algorithms need not be able to affect everyone’s decision. Influencing a large enough fraction of undecided people can be enough to improve profits notably - or swing elections.
Whether or not the data and algorithms were already accurate enough to swing the election for Trump is irrelevant. There’s no reason to expect the rich, politically motivated people behind the scheme to step back now, and hence we should expect more of the same, only with improved performance.
Ok then, what should we do? I see two important directions here: we need to start valuing our personal data much more, and we need to demand more ethical approach from service providers processing our personal data.
Your personal data is invaluable
The first thing anyone of us can do right now is to stop and think about how much and how detailed data and information we share on social media. There are tools that makes this more concrete for us. For example, Data Selfie makes visible the data traces we create on Facebook.
Image adapted from Data Selfie.
PersonalData.IO is a service that helps people to request a copy of the information an organisation holds about them. With the service, created by Paul-Olivier Dehaye, you can equest your information from for example rFacebook or Cambridge Analytica (guide here). Dehaye has even suggested a plan for US citizens to access their data from Cambridge Analytica, which might help shed light on what kind of data the company actually holds. I also recommend listening to this great podcast where Dehaye was interviewed on the topic.
Personal data processing needs transparency and accountability
In the big picture it’s not enough that a group of educated people start controlling their online behaviour, or leave Facebook for good - they were not the targets in the first place. So how to make everyone less susceptible to algorithmic influences?
Obviously, Facebook plays a central role in this, and it has indeed shown interest to address some of the problems raised recently. But I would not count on Facebook only. We must demand the organisations operating on personal data to adhere to current regulation, and to increase accountability and transparency on their actions. In addition, the policy makers and legislators need to keep up with technical development and update regulation when necessary to protect digital human rights and privacy.
To facilitate this discussion, we (EDIT: including Paul-Olivier Dehaye whom I mentioned above) are organising a track on Ethical Processing of Personal Data in the forthcoming MyData2017 conference. The call for proposals is now open, please contribute! See here for details of the different topic tracks, and here for a brief overview of MyData.
Image adapted from MyData2017.
Machine learning Professor Neil Lawrence warned us about the concentration of personal data in the hands of “digital oligarch” already two years ago. It’s about time to take note!
In this post use “algorithm” to refer to a machine learning algorithm that learns from data and produces personalised content. ↩