 It's great that you all came here today as I will need you all as participants in a small experiment to illustrate my research. For the experiment already recruited two volunteers, please come on stage. So imagine, both of you, you are politicians and you want to become the new mayor of Barcelona. For this, you came to this university today to give a political speech. And all of you, you're the audience. You came here to listen to our politicians and applaud if you like what they said. So the good thing about being a politician is you don't have to write these speeches yourself. I wrote them for you. Great. First, we have our male politician giving his speech. And remember, if you like what he said, please applaud. Men and women should earn the same. Great. Thank you. Next, we have our female politician giving her speech. I have two questions for you. First question. Who of the two politicians received more feedback? Our female politician, right? Second question. This is the important one. Who of the two do you think will feel more encouraged to bring up gender issues again at the next debate? My research suggests our female politician. But before I explain this in more detail, a big round of applause for the two volunteers. So why the female politician? Because of a phenomenon called reinforcement learning. Reinforcement learning simply means if you do one thing and you get positive feedback for it, you'll feel encouraged to do it again. And since our female politician received more positive feedback, she'll be more encouraged to do it again in the future. And this is exactly the hypothesis I tested using social media data from Twitter. I downloaded more than one million tweets from members of parliament from all over Spain. To find out which of all those tweets were on gender issues, I used an unsupervised machine learning algorithm which analyzes a text and on its own classifies it into different topics. Looking at them, we can identify topics such as economics, social issues, elections and gender issues. Focusing on those tweets on gender issues, I can do a quantitative statistical analysis to find out if, A, female politicians receive more positive feedback for talking about gender issues, and B, does this lead to reinforcement learning, such that female politicians will be more likely to bring them up again in the future. Let me show you how this works in practice. We have a politician called Ada. Ada has a Twitter account and she has some people following her on Twitter. Sometimes Ada writes a tweet on gender issues and then her followers will react to her by sending her comments, retweets and likes. What the statistical analysis does now is that it tests if Ada receives more positive feedback, will she be more likely to talk about gender issues again in the future? What I'm trying to tell with this is that social media feedback can reinforce stereotypical behavior amongst politicians. Imagine a male politician tweeting on gender issues and receiving zero feedback for it, he will not do it again. So why does this matter for you? Because you and me, we all, we are the audience, we are the citizens, we are the people who give feedback to politicians on social media. So maybe you should be more careful about how we react to them, as it seems they are actually listening.