Harvard Criticizes Google’s Adult Content Filter

Google’s SafeSearch porn filter was found to exclude non-porn sites such as the American Library Association, in a recent test conducted by the Harvard Law School’s Berkman Center for Internet & Society.

Google admits the porn filter isn’t perfect but says the test wasn’t exactly perfect, either. The porn filter, when engaged, automatically excludes any Web sites Google knows about only through link analysis, rather than by actually visiting the site.

For example, if a site prohibits search engines from spidering it, Google might still return a link, as it sees links to that site on other pages. However, since it hasn’t actually crawled the site itself, Google excludes the site when SafeSearch is switched on, because it doesn’t know if porn content is present. Further discussion from Google, as well as comments on the report from search engine marketers, can be found in a recent thread on the issue on WebmasterWorld.com.

Sadly, the Harvard report doesn’t compare Google’s filtering to filtering performed by other search engines. Google might be performing better or worse than its competitors, but that’s not covered.

Harvard’s focus is primarily on how non-porn sites might be accidentally blocked. It’s not an examination of how well the porn filter works to keep out explicit content. This is addressed only briefly. The report says even when the porn filter is engaged, it isn’t 100 percent successful.

Overall, the porn filter comes off sounding like a terrible thing to use. Yet some real-world perspective is in order. I remember writing about the first porn filters for crawlers back in 1998. They weren’t perfect, but many readers were happy to have them. My article from that time began with an example of a teacher who did a search for something innocent only to get porn sites showing up in front of her entire class.

Flash-forward to today. I searched for “dolls for girls” on Google without the porn filter. Among the top listings was “Blood Dolls: Gothic Girls in erotic, nude, & fetish photos.” That’s not something you’d wish your seven-year-old daughter to see. With SafeSearch’s “strict filtering” switched on, the site disappeared (strict filtering means filtering of both Web and image results. By default, “moderate” image filtering is always on).

As I said back in 1998, if you are trying to conduct proper research and are afraid the results might return porn, push your kids out of the room and don’t engage any search engine’s porn filter. If you must search with children present and want some protection, the porn filters will provide some options. As Harvard’s recent findings aptly point out, the trade-off is some important, non-porn sites might get accidentally filtered.

For more help, Search Engine Watch maintains a list of kid-friendly search engines, as well as a guide to engaging porn filters at search engines.

Don’t miss ClickZ’s Weblog Business Strategies in Boston, June 9-10.

Related reading

site search hp
amp
Space Shuttle Launch
google-changes
<