German AI chatbot seeks to make it easier for voters to decide

This weekend, around 60 million Germans are being called to vote. However, according to surveys, around 30% are still undecided as to who they should vote for on Sunday.

Tools such as Wahl-O-Mat, which provides users with a political quiz before giving an answer at the end, are particularly helpful for them. They help in ascertaining which political parties and their corresponding programmes best align with their own values. In addition to Wahl-O-Mat, there is now another way to find out about government programmes, but without having to read hundreds of pages.

Wahl.chat is an AI-supported chatbot developed by a group of German doctoral students at the renowned British University of Cambridge. Their goal: to make politics more accessible and provide voters with simple guidance.

Euronews spoke to Michel Schimpf, one of the developers, about how to make a political chatbot apolitical and how to get voters interested in politics.

How did the development of the chatbot come about?

We are a five-member German team that came to Cambridge University to conduct research together. Like many others at the moment, we probably talk a lot about politics over lunch. We thought about how we could make a small contribution to that.

Then we came up with the election and just started programming. We launched the bot just under a month ago. Since then, everything has gone pretty quickly.

How does the chatbot work?

If users have questions about a particular party and want to know: “Okay, what does party X say about education policy now, for example?” Then you can go to the chatbot and select the party and ask about it.

Other examples can be, what do you want to do for better education? Or, sow do you want to boost the economy? And then our chatbot gives an answer based on the parties’ election programmes.

You can also click on several parties and compare them.

If you tried to open five election programmes at the same time, for example, it would take quite a lot of time. With our chatbot, you can simply ask about the topics that interest you personally.

So the aim was to make decision-making easier for potential voters?

Exactly. In general, we are trying to make politics a bit more transparent.

How was the chatbot generally received?

Very well. Of course, you also get feedback and some things don’t work that well yet.

Generally, we can say that the response is really extremely good.

We also hear from teachers, for example, that are using the tool during their classes to prepare for the election with their pupils. We also hear from politicians who are using it to better understand the election programmes of other parties.

Some even use it to better understand their own election programmes. Given that such documents can be 100 pages long or more, it’s not that easy.

ADVERTISEMENT

What is the long-term goal?

We are incorporating more and more sources. We will also include the speeches made by politicians recently in the chatbot. There are simply so many sources of political information and virtually nobody can keep track of all of them, even people who do this full-time.

We want to create a way for people to find out about politics in a more personalised and easier way.

What is your relationship with AI in general?

Well, I’m doing a PHD on AI in psychology and all the people who are in the team with me are doing research with me. And we have also used AI before, in very critical areas such as psychological interventions. We therefore already have a lot of experience.

What are the potential applications of AI in political decision-making?

There are plenty of possibilities.

ADVERTISEMENT

The most important thing is to create transparency. The Bundestag speeches, for example, are all transcribed and there is a huge amount of information and certainly a lot of valuable things in there.

Who has the time to read through all that, though? With our chatbot, it will be much easier at some point.

Isn’t it a difficult goal to keep a chatbot 100% objective?

That’s true. But you also have to say, what are the alternatives? News isn’t always completely objective either – so if you go on Facebook or X, they’re not objective.

Overall, you always need your own common sense. You have to see what the sources look like. We make that possible, too.

ADVERTISEMENT

With us, you can click directly on the sources and see what’s really in the election programmes.

This is a problem with AI, which we certainly have, but I think it will be much less of a problem than with other sources.

If the bot was biased, that would be problematic. How do you deal with this in the programming?

Our AI is basically instructed not to use any judgemental adjectives and to work based on sources. And then you can get to the source with one click. There is research on this.

The models partly adopt the characteristics of the underlying data, but it’s very difficult to say at the moment if something is more left-wing or more right-wing.

ADVERTISEMENT

The model is instructed to be as objective as possible. I think that works very well.

If you try it out, it’s very rational. In my opinion, you could make it a lot more sensationalised. I think it works well the way it is now.

What does the future hold for your chatbot?

The AI models that already exist are good enough to do almost anything with. We don’t feel limited by that.

The only problem is that it’s extremely expensive. I think you can see on our website that it costs us around €250 a day to run. But that’s a bit out of date. The costs are now closer to €500. Some days, it’s even more expensive. That’s our biggest problem right now.

ADVERTISEMENT

Should the model be expanded beyond Germany?

We’re giving it a lot of thought. People have already written to me saying: “I came to Germany from country X and I think it would be quite relevant in my country, too.”

There are also people who have asked us about translation into other languages. We think that a tool like this chatbot is definitely needed in democracy. You can do a lot of good with it.

That’s why we can imagine continuing with it.