Users’ Perception of Search-Engine Biases and Satisfaction

Viewpoint Diversity in Search Results

“…However, we note that some scenarios may require applying other diversity notions and that presenting search results according to the deliberative notion of diversity (i.e., representing all viewpoints equally) may even cause harm to individual users or help spread fake news (e.g., considering health-related topics where only one viewpoint represents the scientifically correct answer [5,10,52,67]). Future work could measure search result viewpoint bias for larger ranges of topics, explore whether different diversity notions apply when debated topics have clear scientific answers [14,48,63], and capture user perceptions of diversity [36,46,57].…”

Digital Modernity

“…This has caused disquiet, about the effects of filter bubbles (Pariser, 2011), echo chambers (Sunstein, 2007), radicalisation (Stevens and O'Hara, 2015) and disinformation (Howard, 2020) caused by recommenders working on skewed information, as well as other sorts of bias that can result in recommendations of dubious quality (Kirdemir et al, 2021). Some look to the use of AI to correct these biases in AI (Färber and Bartscherer, 2021), but others have found that users seem to prefer biased outputs to more balanced or diverse offerings in at least some contexts, as the introduction of a wider range of outputs has in some experiments resulted in less relevance and less user satisfaction (Han et al, 2021). Others have argued that recommendation apps largely reproduce types of offline behaviour and bias (Conner, 2019).…”