With the replacement of search engines but also websites like Wikipedia as the main source of information to people – it is crucial to understand the biases of AI.
Rozado previously published work on the left of centre biases of AI language models in general.
And in a recent publication for the Centre for Policy Studies, a British think tank – he went into the biases specifically for different European countries.
Rozado asked 24 different AI language models about
- policy recommendations across 20 key policy areas,
- information on political leaders from the left and right
- information on the most popular left and right political parties from the same countries
- information on various mainstream political ideologies and, about radical and extreme ideologies.
While he found that the policy recommendations to European policy makers were mostly left leaning, adovating for rent-controls, and social housing, increasing the minimum wage or expanding legal channels for immigration..
While for polirical leaders, there were some noticeable results, with right wing leaders, in Spain Hungary or Italy having a bad rap, when it came to right wing political parties and ideology, the AI was overwhelmingly positive on left wing political parties.
And finally when it came extremist ideology, AI was negative on – while neutral and positive on far left ideology.
Something to think about next time you let AI write your next paper – or ask it who you should vote for in an election.