Pixel Envy

Written by Nick Heer.

Unbundled

Remember that graphic that circulated many years ago, showing the internet as bundled tiers? It drove home the stakes of net neutrality. Nothing inspires hatred quite so much as cable companies, and seeing a mockup of an internet that runs closer to their model is frightening.

Dieter Bohn thinks we’re reaching a point where that graphic is becoming a reality. Not because of ISPs shaking down Netflix, but because of “assistants” like Siri and Cortana:

These intelligent assistants are great. I use them every day and expect I will continue to use them for, well, ever. But there’s a problem that’s built into them: they only seem to work with certain parts of the web and — here’s the real rub — certain apps. […]

You can get Yelp results in Siri, OpenTable in Google, TuneIn radio from Alexa. But you can’t get everything, fairly and transparently ranked, the way that Google changed search on the web.

It’s a compelling first read, comparing the way search is expected to function with how it actually works in proprietary assistant software. Moreover, comparing it to discussions about net neturality — as Bohn does a little in the article — is an argument I hadn’t considered.

It would be great if there were a way for third-party developers to integrate with personal assistants. I’d love to be able to ask Siri “what’s interesting around me?” and see results from Inquire, or tell it to “play my weekly discovery playlist” and have Spotify start. There are, of course a host of UI considerations — deciding how an intelligent assistant knows which app to use for a given phrase, for example — but it would be pretty rad if it all worked right.

Upon second read, though, Bohn’s argument is, quite simply, bullshit.

Comparing assistant software to Google search? Is Bohn serious? Their search algorithms are anything but “fair and transparent”. You know: their top secret, frequently-changing algorithms that they won’t patent because then Google would have to disclose how they work? Yeah.

Bohn digs deeper into this angle:

That’s all great, but did you know that there’s no universal way for app developers to make their apps’ content readable to every company? Instead, each app maker has to create an index that Google can read, an index that Apple can read, another one for Microsoft, and so on. I’ve been told by people at multiple companies that we’re getting closer to a universal standard that will make creating these indexes less of a burden on developers, but we’re not there yet.

[…]

Web searches like Google and Bing claim to be based on impartial algorithms, not backroom deals. But the way that OpenTable and Yelp and Hotel listings appear in Siri and Google Now is much more opaque.

Okay, having to appease different companies with proprietary stuff sucks. Got it.

But remember Google’s failed “authorship” experiment, which was supposed to give a rankings boost to websites that linked their authors’ Google Plus profiles within the site’s source?

Or what about their current pitch to business owners? They’re encouraged to set up a Google Local Business page — basically, a Google Plus profile, but with business-ey things like hours of operation and industry. The business is then ranked against their competitors in that weird local search results box based on a bunch of signals, such as their Google Plus activity and how long they’ve had a Google profile. How much weight these and other factors have is, of course, a complete mystery because Google is even more cagey about how they rank local listings than they are about their website rankings.

Bohn:

With Google’s Pagerank, there’s at least a nominal sense that users are picking the winners and losers. I couldn’t tell you what makes app results appear inside these assistants. So far, nobody’s saying publicly how their apps figure out what to show you. I’d like to trust that none of these companies are picking favorites based on backroom deals (the kind of paid placement that led people to distrust the AltaVista search engine 15 years ago), but I’d much rather just know how Siri and Alexa and Cortana make those choices.

Google may not be picking favourites based on third-party backroom deals, but they sure were based on backroom deals internally:

One way Google favored its own results was to change its ranking criteria. Google typically ranks sites based on measures like the number of links that point to a site, or how often users click on the site in search results.

But Marissa Mayer, who was then a Google vice president, said Google didn’t use click-through rates to determine the ranking for its own specialized-search sites, because they would rank too low, according to the staff report. […]

Instead, Google would “automatically boost” its own sites for certain specialized searches that otherwise would favor rivals, the FTC found. If a comparison-shopping site was supposed to rank highly, Google Product Search was placed above it. When Yelp was deemed relevant to a user’s search query, Google Local would pop up on top of the results page, the staff wrote.

There’s more, too. Google says that you can’t buy your way to the top of their rankings, but some experts believe that domains with Google AdWords accounts get a little boost. Again, we’re dealing with Google’s “fair and transparent” ranking system here, so who knows?

Bohn concludes:

… there’s a very real chance that these bots are going to become our primary interface to the internet, the medium through which we get our information and our sweet, sweet content. So I think we should demand that the companies that helped us fight the ISP bundle don’t end up becoming purveyors of bundles themselves. We should hold them to the standard of openness that made the web so vibrant in the first place.

Sure, but why not regulate search engines at the same time? Google web search represents a gigantic share of U.S. search engine traffic, and they were found to abuse their monopoly position in anticompetitive ways. Search engines are most people’s primary interface for the web today; don’t these arguments apply similarly to them, too? I’d argue that they do. But I also wish to stress that advocating for the regulations of intelligent assistants by holding them to the standards of Google’s web search is disingenuous and callous.