Elections and Artificial Intelligence: Where We Are and Where We're Going
Disclaimer: This is going to be as nonpartisan and unbiased as possible. However, providing an unbiased take and considering both sides of an argument are not the same thing.
How Can AI Hurt Elections?
Social Media Manipulation: We already saw what happens when real people (ex. Cambridge Analytica) try to use social media to influence elections. Part of that relies on AI for demographic population targeting, which groups people on social media based on their interests and tries to predict what they might want to see based on that.
Cyberattacks: This is also usually bots, performing automated hacks like phishing and password cracking on political campaigns
Politcal Bots/Fake Accounts: Bots used to be easier to identify and take down, but AI (specificaly NLP) has made it a lot harder to determine what is a person and what is not. This has not been used to influence elections yet, but we’re probably not far off, and non-AI bots are already getting pretty good at engaging in conversation without revealing that they’re bots (find examples). Largely relies on open source programs (which are available to the public), so the technology definitely exists. These bots have also been used to spread right-wing propoganda and fake news on social media on Twitter, Facebook, and Reddit. (Link in the Description for the bots that have been used)
Deepfakes: Deepfakes are videos generated by machine learning algorithms that show someone doing or saying something they didn’t actually do. Sometimes they can be fun (show everybody dance now) but they can also be incredibly dangerous (show the obama ones). They also aren’t hard to make - I personally think it’s great that open sourcing code is becoming a norm in machine learning, but it does mean that algorithms that do this are pretty easy to come by with little expertise needed to make them work. Senator Mark Warner wrote a white paper on regulating social media that described deep fakes as “poised to usher in an unprecedented wave of false or defamatory content.” So far, we don’t know of any instances where deep fakes have been used for political campaigns in the US, but a Belgian socialist party has used deepfakes for campaign messaging (show video) by making videos of Trump taunting Belgium for not meeting the conditions of the Paris Climate agreement (which is… odd). The video shows some clear distortion, and the actor voicing the video says that it is fake at the end, but many people in the comments say they thought it was a real video. On a positive note, there is research going into detecting deep fakes. DARPA has reported positive results by watching for unnatural blinking in videos.
AI Politicians: An AI bot that ran as a candidate in the recent Russian presidential elections and was created by Yandex, the Russian equivalent of Google. Only received 25,000 votes. Similarly Michihito Matsuda ran for Mayor of Tokyo, and was third with 4,000 votes. SAM in New Zealand is running in the 2020 general elections. This brings up many questions, including: what happens if an AI politician is elected? can an AI political be appointed by a person? Who has control over an AI politician? How do we handle the ethical issues of AI? There’s a lot of potential pros and cons to an AI politician, and Abishur Prakash summarizes them well in this medium article (https://medium.com/politics-ai/ai-politicians-a-revolution-in-politics-11a7e4ce90b0), so if you’re curious, check it out.
AI-Based Political Chatbots: Have someone to answer voter questions 24/7, whether you’re curious about a party’s stance on an issue or are canvassing and don’t know an answer off the top of your head.
How Can AI Help Elections?
Fighting Disinformation: AI has been used to fact check articles and posts (Factmata and Avantgarde Analytics) and debunk known falsehoods. It can also be used to educate voters on political issues. Finally, it can be used to disambiguate information collected from an electorate so that representatives can work on important issues that may be unseen otherwise.
Online Elections and/or Non-Voter ID ways of identifying people
Face Verify by Machine Box - Person uploads their info via drivers license and then uses facial recognition to decide whether you are the person you said you are. Cons of this are that most webcams don’t recognize depth because you need two cameras, so this would be easy to cheat. Similarly, things like FaceID
Media Lab Proposal (Cesar Hidalgo, Collective Learning) - Automated system that collects our interests, views, etc from social media, and then votes on our behalf - Removes politicians from the equation.
Lots of problems here: Security, voter disenfranchisement via lack of access to technology, not everyone is on social media, who proposes bills, how does the AI know when you’ve changed your opinions, how does one advocate for things, etc.