A federal judge threw out KinderStart’s sweeping lawsuit against Google yesterday, but left the door open for the whole thing to start again. The key thing here, I think, is the concept of “objectivity”. From ZDNet coverage:
(Judge) Fogel wrote that a PageRank score reflects Google’s opinions, and suggested that a score generated by a computer algorithm is likely not defamatory. But if KinderStart can prove a case of “manual intervention” by Google, Fogel wrote, the outcome might be different.
Innaresting. So as I wrote last week, it appears that the subjective selection of features (factors) is ok. But, once those subjective features are opinionatedly chosen, only the changes in objective values of those subjectively selected features may be used to determine rank…otherwise Kinderstart may have a case.
The problem is: How does Kinderstart actually go about proving “manual intervention”? Assuming from our earlier discussion that the lawyer’s comments last week were not an actual admission of such (I still think they were an admission, but others here have made quite good points), the only other place I can think to start is, again, that Wired magazine article from three years ago, where it really does sound like Google admitting to hard coding/manual intervention:
Finally [the Google person] stops and looks up at me. “They’re going to be OK. They got referred to the right places.”
“You can do that?”
“Yeah, well, I can see how the system responds. And if it doesn’t give the right information, I’ll find better sites and attach them for future queries.” [emphasis mine]
And while that might not prove anything directly about the current case, it at least opens the door by showing Google’s willingness to “manually intervene” even in ways they’ve otherwise publicly stated that they don’t.
I talked with Google after that article came out. They said the engineer meant coming up with an algorithmic response. See http://blog.searchenginewatch.com/blog/051012-110515
Ah, yes, Danny, thank you. This has been bothering me for.. well quite literally.. three years. In this time, I’ve actually asked numerous Google engineers about it, and not gotten a satisfying response as to why that exact language was used.
The response I usually get is the same one you’ve gotten: I was told by Google that this was a misinterpretation on the part of Wired. The Google engineer apparently meant that the Google search algorithm would be tweaked to produce better results, not that the results would be reordered by hand.
However, I’m still uneasy about this because it doesn’t quite add up. Google claims it was a “misinterpretation” on Wired’s part. But I see quotation marks around the Google engineer’s words. I see a Google engineer saying “I’ll find better sites and attach them for future queries,” with no commentary or interpretation (much less misinterpretation) from Wired.
Maybe the engineer was trying to dumb it down for the reporter. That’s always a possibility. But I’ve never heard a serious information retrieval researcher talk about search in those terms. It is always “I’ll find a parameter setting that will improve the results”, or “I’ll do some better feature selection that will lead to an increase in (relevance) precision”. Never “I will find better sites and attach them for future queries”.
Do you see the difference in mindset and thinking in these responses? Someone who is really twiddling with algorithms speaks in terms of features and weights. That is not the tone or language of this engineer. At all.
The problem is that, no matter what the Google engineer actually meant, Google PR is going to tow the company line and say that he meant an “algorithmic response”. So it makes it difficult to really know.
My own personal opinion is that we managed to catch some Google engineer in a moment of rare frankness or openness. But of course I’ve got nothing to back that up.
The comment disturbed me as well, which is why I specifically tracked it down with them. I agree, it could entirely be after the fact spin. But we’ve really never seen any strong evidence of Google hand crafting results, as we have had with some other search engines. So I tend to believe it was the engineer either not being clear, the reporter not getting it down exactly right (both are common) or a combination of them both.
Danny, I’m wondering if you actually were able to speak with the Googler in question, the one that issued the quote. It would be interesting to get his perspective on what he actually meant. Well, again, I know what he would now tell us that he meant (“I mean, algorithms, guys…honest!”). But it would be useful to be able to ask him why he chose that language, or if the reporter misquoted him.