What role does AI play in the election campaign?
To what extent does artificial intelligence influence statements by politicians? Now and in the future? What is meant honestly and what is not? Our guest author gives a few thoughts to the point.
The past weeks and months have shown very clearly how decision-making processes work within parties. At least when it comes to personal details. In the discussion, terms and expressions such as majority polls, mood in the electorate, what or whom the members prefer, and so on were often used. Before every election there is a regular Sunday question. On the evening of the election, the obligatory phrases such as “calmly analyze the result” are used in the elephant rounds – mostly from the losers of an election.
Contents
Analyze what exactly now?
Surveys are usually carried out as they were tens of years ago: via telephone interviews. To this end, people are called and asked to give supported answers. So whether this or that corresponds to the personal opinion. In contrast, market researchers call a free choice of statement “unsupported”. The principle is known from quiz programs: For some, an answer has to be given directly, formulated yourself, for others, candidates have the choice between two, three or four answers. Examples are the “family duel” (if you still know it), in which the task was roughly: “Name things that people think of about the weekend”; and the classic “Who Wants to Be a Millionaire?” – for which there are four possible answers. Political surveys usually follow the latter pattern.
More than you might think, AI can change things massively here – in terms of moods and decisions. From a purely practical point of view, the initial situation looks like this: surveys take time. It just takes time to interview thousands of people personally. Supported digital surveys are of course faster and can be evaluated in real time. But they have to be initiated and controlled. That actually means a distorted picture. If a vegetarian is given the choice between schnitzel or steak, what will the answer be? A trained, self-learning AI can deliver far more, faster and more meaningful survey results – with NLP, natural language processing.
As mentioned, traditional surveys are tedious and the evaluation results are one-dimensional. Now purely theoretically – whether this is already being used is beyond any knowledge – considered: an AI can recognize and analyze text on social media channels and in forums and blogs. The original statement, the topic and the answers available to it. The text engine recognizes the meanings of words and contexts (only as an example: “Bank” for sitting or “Bank” for “payment transactions”). In addition, an AI can also identify sentiments and then cluster them – called “sentiment analysis”. So whether a comment is benevolent or critical.
This can be done within seconds – with lots and lots of data. What is generally known is that in parties like in large companies there are entire departments in which this is done manually. There are then a dozen or more people at the computer and all of them go through websites, blogs and so on individually and note, evaluate and classify the results – according to specifications and / or personal opinion. It behaves like the classic PR clipping familiar from breakfast television: a summary of which daily newspaper reports what and how. And that is exactly what an AI can do – only in a different and much larger dimension.
Ask yourself: How many websites, blogs and social media posts can 20 employees manually go through every day? Read everything and classify? And then analyze it? An AI can do this with a power of XX. Real time. Admittedly, a human employee can (even) assess this better. But developments and advances in this area of AI are rapid. And through the mass analysis, very clear mood images can definitely be determined.
Also interesting: The AI privacy dilemma – can we use powerful artificial intelligence and still protect our privacy?
The gloomy picture that emerges is that, for example, a politician gives an interview on TV in the morning and can see just a few minutes later how it has been received. Which statements were commented on and how? What was well received? What was rated positively? What are the criticisms? With the answers, the politician can already respond concretely in the midday magazine. It is better not to say that, to formulate it differently, etc. It may appear gloomy if one assumes that politicians only say what goes down well. No longer what they actually stand for. But you can also look at it positively.
Politicians and parties are often accused of being unrealistic. There are enough suspension points for this. If, for example, an economic politician cannot answer the simple question of what a pound of butter costs in the supermarket – not good. By means of AI, politicians – people and parties – receive precise and detailed impressions of what is going on in the population very quickly. Apart from the (mostly the same) respondents from opinion research institutes. And politics – people and bodies – can of course also take this into account when developing election manifestos.
But you should also ask yourself two questions, since AI ultimately only evaluates data, does not form opinions itself: Who comments in forums, predominantly on social media? Is that trending or tendentious? The answer to the first question is still relatively simple: Everyone comments, left and right and middle, young and old. Everyone has to find the answer to the second question for himself. Just like politicians have to think about why they say something and what they publicly stand for. Out of conviction or based on opinions determined by means of AI?
