Exploring acquiescence response bias using digital footprints

Acquiescence response bias is the tendency to agree to personality questionnaires irrespective of item content or direction. Acquiescence is problematic for both researchers and clinicians (Rammstedt & Farmer, 2013), and yet is mostly ignored or treated as meaningless error. While acquiescence is known to be shaped by both age and education (Costello, 2015), little more is known about the behavioural implications of acquiescence. Using a sample of 20899 US adults from the myPersonality project who completed the 100-item IPIP inventory, the current study explored the extent to which social media behaviours (Facebook likes) could predict acquiescence using machine learning. Acquiescence scores were calculated using matched item pairs (Hofstee, ten Berge & Hendriks, 1998). Using singular value decomposition (SVD), linear regression, and cross validation, the optimal prediction accuracy with 27 SVD dimensions was .31. This suggests that acquiescence is multidimensional, can be predicted using digital behaviours, and therefore likely plays a role in everyday behaviour. Implications for both researchers and clinicians are explored.