Why did two people's Google search for "BP" look nothing alike?

Because Google was personalizing them. Personalized search results are based on information about the user, such as location, past clicks and search history, so two people typing the same words can get different pages. The internet activist Eli Pariser, who coined the term "filter bubble" around 2010, described one user who searched "BP" and got investment news, while another got information about the Deepwater Horizon oil spill: two results pages he called "strikingly different." When he asked friends to search "Egypt," one friend's results prominently featured the 2011 revolution then under way, and the other's first page didn't. A Google engineer told him the company looked at 57 pieces of data, including computer type and location, to tailor results.

Pariser's 2011 book The Filter Bubble predicted that this personalization would leave people intellectually isolated, with less exposure to conflicting viewpoints. Many people don't notice the curation at all: more than 60% of Facebook users were found to be entirely unaware of any curation of their feed. A Facebook data science study found that the feed algorithm reduced politically cross-cutting content by 5 percent for conservatives and 8 percent for liberals, while users' own choices reduced their clicks on such links by 17 and 6 percent.

How much this matters is disputed. There are conflicting reports about how much personalized filtering actually happens. In a small, non-scientific 2011 test, analyst Jacob Weisberg had five people with different politics run the same searches and found results that varied only in minor ways, and one study concluded that political polarization is driven mainly by pre-existing views, even though algorithms do weaken the diversity of what people see. Pariser saw a threat to civic discourse; contrasting views regard the effect as minimal and addressable.