algorithmic bias in google searches for political parties
play

Algorithmic Bias in Google Searches for Political Parties and - PowerPoint PPT Presentation

Algorithmic Bias in Google Searches for Political Parties and Candidates Johannes Nakayama, Nils Plettenberg, Laura Burbach, Andr Calero Valdez Human-Computer Interaction Center, RWTH Aachen University The Datenspende BTW17 Dataset search


  1. Algorithmic Bias in Google Searches for Political Parties and Candidates Johannes Nakayama, Nils Plettenberg, Laura Burbach, André Calero Valdez Human-Computer Interaction Center, RWTH Aachen University

  2. The Datenspende BTW17 Dataset

  3. search terms search terms … database search terms Ø 1647 searches per day 4000 participants

  4. parties: candidates: ‘CDU’ ‘Angela Merkel’ ‘CSU’ ‘Martin Schulz’ ‘SPD’ ‘Christian Lindner’ ‘FDP’ ‘Katrin Göring-Eckardt’ ‘Bündnis 90/Die Grünen’ ‘Cem Özdemir’ ‘Die Linke‘ ‘Sahra Wagenknecht’ ‘AfD’ ‘Dietmar Bartsch’ ‘Alice Weidel’ ‘Alexander Gauland’

  5. Can we find evidence for personalization in the dataset?

  6. First approach set intersection: first three/six/nine results

  7. A C B D C A D G E F F B G H H E I I

  8. A C B D C A D G E F F B G H H E I I

  9. A C B D C A overlap of 66.67 % repeat for every list with every other list compute mean

  10. downside: only rough order effects

  11. Rank-biased overlap (RBO) (Webber et al. 2010) Result 1 Result 1 Result 2 Result 2 Result 3 Result 3 Result 4 Result 4 Result 5 Result 5 Result 6 Result 6 Result 7 Result 7 Result 8 Result 8

  12. How does RBO work? Identical Lists Entirely different Same items, reverse items order User 1 User 2 User 1 User 2 User 1 User 2 A A A K A J B B B L B I C C C M C H D D D N D G E E E O E F F F F P F E G G G Q G D H H H R H C I I I S I B J J J T J A RBO score: 1.0 RBO score: 0 RBO score: 0.5116076

  13. Last 5 items different How does RBO work? User 1 User 2 A A B B 1st item different 3rd item different C C User 1 User 2 User 1 User 2 D D A K A A E E B B B B F F C C C K G G D D D D H H E E E E I I F F F F J J G G G G K P H H H H L Q I I I I M R J J J J N S O T RBO score: 0.7297158 RBO score: 0.8747158 RBO score: 0.901564

  14. Similarity of Search Results

  15. T-test significant differences between RBO scores of parties and candidates t = 27.4, p < .001

  16. H o w c a n w e e x p l a i n t h e d o w n w a r d t r e n d ?

  17. Search Volume no absolute values only relative scores Trends

  18. Summary significant difference between parties and candidates dependence between RBO scores and search volume

  19. Future Work R T RBO package location content analysis (already in progress)

  20. open data and all analyses available at: osf.io/e598k

Recommend


More recommend