On Algorithmic Bias

16 November 10

In a recent article, Hard-Coding Bias in Google ‘Algorithmic’ Search Results, Benjamin Edelman details what he believes to be hard-coded bias in the manner Google displays results for certain searches:

Search for a stock ticker (example: CSCO), and the three most prominent links on the page — the large-type all-caps ticker symbol, the large price chart, and the left-most details link — will all take you to Google Finance. Google Finance isn’t the most popular finance site; according to ComScore, Yahoo Finance claims that title, and indeed ComScore puts Google Finance in position #60 (as of April 2010). Nonetheless, the three most prominent links all promote Google’s in-house finance service.

Edelman goes on to cover searches for health terms, airport codes, movies, and patents and how each — when searched for using a specific format — result in Google giving preferred placement to its own or partner properties. By this, he comes to the conclusion that Google is abusing it’s power and is in violation of their “we do not manually change results” promise:

Indeed, Google’s use of hard-coding and other adjustments to search results gives Google an important advantage in any sector that requires or benefits from substantial algorithmic search traffic… I am concerned that Google has made inaccurate representations to the public including to users, publishers, advertisers, investors, and regulators. Comparing Google’s “we do not manually change results” and similar promises to Mayer’s quote and my findings, I can only conclude that Google’s promises are not true…

The mistake Edelman has made is in confusing the results that are returned as part of Google’s search algorithms and those that are specific to what Google has determined is the user’s intent; i.e. for what the user is actually searching.

If someone enters a stock symbol, it’s very likely that the current state of the stock is the desired result. Given that Google has their own finance service, it makes sense that they would use it as the source for this data. When Google lacks an internal source for this information, they tap into the services provided by their partners.

From a service standpoint, this is beneficial to Google and their users. By controlling the source of this data, Google is better able to ensure its availability. It wouldn’t make sense for them to rely on the ability other content providers to do the same, especially in the absence of formal partnerships. Users win because the information they were most likely interested in surfaces to the top — no further action required.

In most cases (movies being at least one exception), Google does provide outbound links to other content providers. These links appear to be “hard-coded” to a specific set of trusted, popular, or otherwise well-known sites. Yes, Google’s own properties appear first.

Edelman’s own findings make it is apparent that the display of these search “results” is based on very specific search input formats. I have no insider knowledge of how Google handles this within the larger set of search results; they could be algorithm-based, hard-coded, or a mixture of the two. Regardless, it does appear to be the case that this promoted content is in-fact separate in some manner from the standard search results.

Perhaps Google would do well to make the promoted nature of these content blocks more obvious, but their failure to do so should not be taken as an indication that they breaking any promises or being “evil.” If we are to take Google to task for how they display this content, then other search engines should be as well.

The question is, are users better served or harmed by promoted content within search results when that content has a higher potential to meet their immediate needs?