What do the candidates tell us?
The Chancellor’s duel between Angela Merkel and Martin Schulz for the 2017 federal elections was shown on four channels, with correspondingly many presenters. The stations also allowed themselves to ask the viewers live who had won the duel. The result: very different. The criticism became loud live in social media as well as in the press: major issues of the future such as education and digitisation were completely absent. It was also difficult for four moderators to draw the big bow. What were the opponents talking about in the end?
Artificial intelligence clarifies
Our passion is to bring man and machine into a good relationship. We (or our “software sister” SmartMunk) program text analysis systems that make it possible to examine the texts objectively (machine stop). The aim is to uncover structures so that people can form a better picture. We invite them to do so.
Talk a lot versus say a lot
A first indicator is the amount of speech and the information content of the statements. The moderators had, with their questions, the largest share of speakers (48%), but naturally little information content. The “Content Indicator” (last column) is a measure of the richness of the statements. The more new things are added to the call, the higher this value is.
It is interesting that Martin Schulz had the longer speech (or spoke faster at the same time), but in the more words he did not contribute more information. On the contrary, the Chancellor shows the highest information content in her statements. Apparently she touched on more topics and new fields of conversation than her opponent. Trump and Clinton also showed a similar relationship in their 2016 debate.
More emotion with the chancellor than with the challenger
The results of the text analyses are “maps” that reveal the contents and their structure. We have posted some of these maps for you here. For example, the TreeMaps, in which the speeches are displayed as a visual mapping. The so-called ontology consists of nine categories as a kind of meta-category system: Actions (red), functions (green), emotions (pink), people (orange), places (brown), time (grey), product (turquoise), brand (blue), advertising (purple).
We were astonished at the first comparison: Angela Merkel was more emotional in her statements than her challenger. Most of the speeches were actions (red). We know similar things from analyses of past political duels, such as Trump against Clinton. The Chancellor’s contributions to the debate were based on rational arguments (17% of her contributions), followed by emotional aspects (13%) and people.
Mr Schulz’s relationship is different. After actions and reason follow statements about persons. It is striking that the Chancellor makes much more frequent reference to refugees and names specific occupational groups such as police officers, the unemployed or families, while Martin Schulz remains more general.
The human factor versus Europe
A further view of the statements helps us to better understand the core of the statements and their context. While the Chancellor shows a lot of relation to the human factor, Martin Schulz spoke much more about Europe, its neighbours and corresponding politics. Certainly the result of his past as a European politician.
The analysis is not intended to predict a winner. But it offers a rational view of the debate. In the future, artificial intelligence will increasingly support us in discovering structures where people cannot see them right away – for the abundance of data. The selection of data and interpretation of results will continue to be done by us humans. We are curious what your interpretation of what you found is.