Why should Google be any different than the other filters out there? There are companies whose entire purpose is to correctly filter out the naughty sites, and they unfailingly block sites that are useful (CDC, breast-cancer sites, and so on). Especially in an automated fashion, it’s tough.

That’s one of the reasons that the ACLU and librarians don’t want to have to install filters on library computers: So much good stuff will be blocked out, as well.

While Google is certainly positioned to do a better job than the net filters, I never really imagined that Google would do that much better (see No. 3 for more on this), at least at first.

It’s always good to have watchdogs out there – for causes I believe in, those I don’t, those I don’t care about. This group examination is a good “checks and balances” type system. Good for Ben.

It’s doubtful that any harm will come from this analysis/publicity. Yes, Google may have to work a little harder, but that will earn them some respect etc. We can all win. Good for Ben, again.

I read so much hand-wringing reaction over this “discovery,” mainly on blogs but in tech columns, as well; I just don’t understand the fuss.

Let’s look at a few facts and observations:

• Google never promised that the SafeSearch filter was 100 percent accurate.
• What is 100 percent accurate? I think access to information about contraceptive choices should be allowed through; you may think this is unsuitable for your child.
• Google’s response to this study is that they try to err on the side of caution: Whether or not this is true, it seems to be a good policy. Kind of like the “innocent until proven guilty” concept. If in doubt, suppress. AND NOTE that this suppression is not

censorship. The user turned on the filter, and can always turn it off and resubmit.


• You don’t have to use the filter – Unlike the debate over library filters, Google can be used in two ways: Filtered and unfiltered. Feel like you’re missing things? Turn filter off (this is the default). Getting too many naughty links for your taste? Turn the filter on. Your choice.
• Google is not in the business of this type of filtering – the accuracy of their filter is probably not as high a priority as other projects/tools. Let’s be realistic.(Note: I’m fully aware that Google is, basically, a harvesting and filtering company, so filtering [indexing, page rank etc] is key to its operation. But not in the “naughty or nice” way — at least not currently)

Lee,

Your thinking as to the context of this work, its value, and its reception

in the community and media is generally consistent with my own.

I do disagree somewhat with your suggestion that there was no reason to

think Google might do a far better job in this field than anyone else. They

have quite a talented bunch of engineers, with impressive results in major

fields (i.e. core search quality!). They also have a huge database of

content in their cache. And I, at least, have found it difficult to get a

good sense of just what AI systems can do and what they can’t — on one

hand, they’re clearly still imperfect, but on the other hand I’m sometimes

shocked by just how good they can be. All that’s to say — I started this

project with the sense that SafeSearch might well get a clean bill of

health.

My real focus here, though, did become the “push Google to be better with

this,” as you propose in your #2. The service has been in place for three

years without, I gather, any large-scale investigation of its accuracy or

effectiveness. (And I say that with full readiness to admit that there’s

lots more I, or others, could do; I’m not sure I’d call what I’ve done so

far a “thorough” investigation, given the millions of search terms and

billions of result pages not checked.) I’m hopeful that my work will cause

Google to reevaluate some of their decisions and, perhaps most importantly,

improve their transparency and documentation as to how the system works.

As to the “who cares” reaction — there’s always the potential, in blogspace

as well as in commercial news sites, for a story to get overblown. I’m not

immediately prepared to say whether that’s what’s happening here.

Personally, I think coverage like that on http://dognews.blogspot.com/

(see the 3:31PM post of yesterday; their deep/permanent links unfortunately

aren’t working quite right at present) isn’t such a bad thing and doesn’t

make the world a worse place!

Anyway, thanks for the clear thinking here and the explicit taxonomy of the

several approaches to this project. That’s a nice and, I think, helpful way

to present the varying perspectives here.

Ben Edelman

Berkman Center for Internet & Society

Harvard Law School

http://cyber.law.harvard.edu/edelman