In the paper embedded into this post “Domain Bias in Web Search,” several Microsoft employees and Stanford University researchers seek to establish definitively that a “domain bias” exists on the internet wherein consumers, irrespective of all other factors, will click on a domain name even as all other variables are changed around it. But what makes a domain name trustworthy to an online consumer of information, products or services?
However, clicks are fraught with biases. The most widely studied bias is position bias [12, 13], a user’s propensity to click on a search result just because it appears closer to the top of a search results page. Much work has been invested in both establishing the existence of position bias [8, 18], as well as understanding how to remove position bias from click activity [2, 3, 7]. Other biases are also known to exist, for example, snippet attractiveness bias, a user’s propensity to click on a result because the query terms appear in bold in the title multiple times .
In this paper, we uncover a new phenomenon in click activity that we call domain bias—a user’s propensity to click on a search result because it comes from a reputable domain, as well as their disinclination to click on a result from a domain of unknown or distrustful reputation. The propensity constitutes a bias as it cannot be explained by relevance or positioning of search results.