Twitter shows ND winning with 6 percentage points from SYRIZA. This is the conclusion of the Political Barometer, developed by the Artificial Intelligence and Information Analysis Laboratory of the AUTH Informatics Department using a new political cafe: social networks.

In particular, “Political Barometer” is a software for assessing political public opinion using the analysis of political tweets related to political parties, which are collected and analyzed daily.

“The analysis is carried out using artificial intelligence methods (emotional analysis of texts). This applies to the political parties of the Greek Parliament, for which sufficient data can be automatically collected from Twitter,” notes Mr. Ioannis Pitas, Professor of Computer Science at AUTH and President of the International Academy of Doctoral Studies in Artificial Intelligence.

In particular, according to the Political Barometer, the assessment of the intention to vote relative to the valid ones in Sunday’s elections is 36.3% for New Democracy, 30.1% for SYRIZA, 10% for PASOK, 4.2% for MeRA25, 3.7%. for Hellenic Solution, while 8.8% are collected by other parties. Judging by the analyzes, there are no serious deviations from the poll results yet (from May 2022 to May 2023). The deviation is about 0.5-1% for small and large lots, respectively.

At the same time, of course, the APTH method provides important political information. Revealingly, according to Mr. Peet, the total number of tweets shows that the TV brawl of political leaders had little effect on people active on Twitter. Also, the case of the possible non-candidacy of Mr. Pavlos Polakis surpassed even the tragedy of Tempi in influence on Twitter. On the other hand, it became evident very early on that the political impact of the Tempe tragedy was rather temporary.

Twitter audiences have special characteristics. For example, these are mostly young people, in the comments they tend to be negative, i.e. even come to insults. Some parties (eg SYRIZA) have a larger and comparatively better presence on Twitter compared to their appeal to the electorate as a whole.

A political barometer, with the proper use of past polls and machine learning methods, can successfully narrow down the analysis to the general electorate (more precisely, the pollster) and make appropriate predictions.

Basically, it can replace surveys while offering significant benefits. It is characteristic that the analysis of public opinion is daily, with a direct reaction to political incentives. In addition, the cost is almost zero.

“Even if there are failures, this method is not going anywhere: as long as it is fueled by the right and big data, and as long as people use social networks as a platform for political dialogue, the better it will become,” Mr. Pitas says.

The Political Barometer software was tested for a year. In the absence of objective truth (that is, the results of real elections, since elections were not held last year), he uses as “objective truth” to refer to the total electorate all polls of all companies up to a week ago. While not exactly a polling method, the Political Barometer stopped making predictions on May 13th. He will release his predictions again after the polls close on the evening of 21 May.