The Court of Milan has rejected an appeal of a January 21/25 order requiring Google to suppress automatic search query completion suggestions associating two defamatory keywords, truffa (scam, swindle ) and truffatore (con man, crook) with a person operating in the financial services sector. While Google performs some preventive censorship based on local social and legal norms, this is at least the second time a European court has ordered the removal of phrases related to con artists & swindlers.
Figure 1: Based on a search for the plaintiff’s presumed full name & name + “t” (the last name is mentioned in the judgment), Google appears to have complied with the ruling. The plaintiff argued that Google should have applied filters proactively to avoid injuring a person’s constitutionally guarantied rights (which rights isn’t clear – honor isn’t listed in the Italian constitution) and, once notified of the problem, still took no action. Google argued that suggested queries are automatically generated based on prior user queries and thus Google is not responsible for their content. Google also argued that censoring suggestions would open it up to damage claims from those accusing it of suppressing information available on the web.
Autocomplete: one more search reputation management problem
Google’s autocomplete suggestions attempt to anticipate what a searcher is looking for based on Google’s knowledge of previous search queries, i.e. what combinations of search keywords have been popular in the past. If a company or a professional is caught up in a scandal, even if through no fault of their own, a significant volume of searches for the name together with “defamatory” keywords may ensue as people use Google to discover information about the scandal. A user typing the person’s name may then see “compromising” words associated with the person’s name. The search suggestions represent the current collective zeitgeist as expressed by the search behavior of search engine users. A search on the name of Italian’s prime minister, Silvio Berlusconi, illustrates this phenomena:
|berlusconi ruby||berlusconi ruby|
|berlusconi scandal||berlusconi dimettiti|
|berlusconi women||berlusconi news|
|berlusconi bunga bunga||berlusconi silvio|
|berlusconi prostitute||berlusconi bunga bunga|
|berlusconi news||berlusconi wikipedia|
|berlusconi girl||berlusconi intercettazioni|
|berlusconi wife||berlusconi wiki|
|berlusconi trial||berlusconi e ruby|
Figure 2. Google Autocomplete suggestions for Silvio Berlusconi
Autocomplete not only assists query completion, it also facilitates serendipity searching, the discovery of useful related information based on an initial concept and the questions being asked by others.
For people or companies encountering their name associated with unflattering keywords, Google’s auto complete search suggestions present an additional search reputation management problem. A typical solution used to manage reputation problems in Google search results is to move unflattering results down the page by optimizing more favorable results. Some SEO practitioners have successfully used Amazon’s mechanical turk and other means to seed Google’s query suggestions, but in most cases this won’t be a viable medium to long term solution due to the mechanisms behind Google Suggest and Google’s terms of service. Certainly a court order, if issued in a timely fashion, would be a very effective means to resolve a Google suggest reputation problem.
Google’s defense and the Milan Court verdict on Google Autocompelete
In its primary line of defense, Google relied on EU directive ISP / Hosting safe harbour provisions i.e. as the mere supplier of a neutral platform which hosts data inserted by others, Google isn’t responsible for the query suggestions appearing in search autocomplete. The court determined that this argument doesn’t apply to Google Autocomplete. In the court’s opinion, autocomplete is an optional search facilitator – autocomplete’s elimination or modification would not change the freedom of searchers to find content on the web using Google. Google’s guilt lies in the choice to use its system of autocomplete to facilitate searches. The court noted that “This is a choice which clearly has precise business value. (Si tratta di una scelta che ha chiaramente una valenza commerciale ben precisa)” as if search query suggestions were somehow a unique optional compared to the rest of a search engine’s feature set.
Yet as anyone who has spent any time using multiple search engines knows, there are many features, which taken together, differentiate a mediocre search engine from a great one.
Just consider the frustration of viewing results which are out of date, a problem of recency, or off-topic, a problem of relevancy.
The court decision only applies to the specific defendant. Currently a search on what has been called Italy’s “Madoff”, Gianfranco Lande, does not produce suggestions including truffa. Yet a search for the failed chain Aiazzone for does.
Information discovery: why search engines exist
The very reason search engine technology exists is to assist users in a discovery process which leads to finding a needle or two in the vast haystack that is today’s world wide web. Information discovery relies on success in each of the tasks in the information retrieval (IR) process: the crawling, indexing and querying of web content (see Andrei Broder’s classic A taxonomy of web search for a more in depth look at how & why people use search engines). Over the years Google has refined its query interface to further promote information discovery – from Google Instant results to search query spelling correction and site link suggestions. Autocomplete, first launched as a test in 2004, is just one of Google’s ongoing search enhancements, most focused on facilitating the rapid and efficient discovery of information.
The court’s claim that Google’s autocomplete is somehow a distinct optional feature enriching Google differently from other elements of Google’s information discovery system defies logic.
Shooting the messenger: the court’s censorship of information discovery
By forcing Google to censor the serendipity search part of the information discovery process, the court makes it much more difficult for potential clients of the plaintiff to discover if he is indeed the same person who appears to be pitching what might be thought of as a get rich quick scheme. That the two share the same last name may be a mere coincidence, but it is imperative that people are able to make an informed decision before spending or investing their money – such is their right too, a right the court has trod upon with this decision.
The court also said that Google suggest is only abstractly neutral as such neutrality is lost when an automatic algorithm produces “improper” keyword combinations. Yet these keyword combinations reflect nothing more than what search engine users type while performing their due diligence on a person who purports to be a financial services specialist. If there are “improper” keywords associated with a person’s name, such is just a transparent reflection of a person with that name having a wider reputation problem.
Should the plaintiff be engaged in questionable conduct, as many people seem to suspect based on their search queries, the court’s decision effectively aids and abets such conduct by obstructing users in their due diligence information discovery process. If a search engine’s mission is to facilitate access to information, what place does the court have in obstructing this facilitation through its censorship?
Google’s missed opportunity?
From a reading of the ruling document it doesn’t appear that Google considered highlighting the fact that when keywords appear next to names in Google autocomplete, there is nothing that specifically links those keywords to a specific person. Many people share the same name.
Plaintiff’s missed opportunities?
A court injunction does seem to be an efficient solution to a search query reputation problem. What is puzzling is that the plaintiff went to all this trouble over just two unflattering keywords. Not only are Google’s suggestions algorithmically generated, they are dynamic, they change over time to reflect current searches. Figure 3: What about a reputation problem in bing? By freeing up two suggestion slots, the plaintiff runs the very real risk that other defamatory keywords will appear next to his name. Why didn’t the plaintiff provide an extensive list of negative keywords? The bigger question, should the plaintiff be the same person people are searching for, is why the plaintiff has a reputation problem, and is the plaintiff willing to address it at the source rather than censoring the symptoms? A court order targeting Google won’t resolve a reputation problem in Bing, or elsewhere, for that matter.
Google’s public response
To date Google has not commented on this ruling on its Italian blog nor on its European public policy blog although it may be telling that Google is hiring two people to work on public policy in Italy.
In related news from Italy, a Rome circuit court recently ruled that Yahoo is responsible for piracy occurring on sites listed in search results.
- Google Autocomplete, née Google Suggest, the Precursor of Google Instant
- Between the Lines: Google’s March 2012 Search Algorithm Updates
- Google removes Suggest result counts – even behind the scenes
- Move over Basic Search Results, Google Rich Snippets are here
- Google Instant Changes Little – and Some SEO Considerations using the Italian ABCs