So I do Google Opinion Rewards to earn Google Play Credits to play Final Fantasy Record Keeper. If you grew up gaming in the late 80’s early 90’s I highly recommend checking it out. Anyway, I was going through a survey this morning and after the first question of “What Types of Searches” have I done this week, it got really really interesting. Google Opinion Rewards are normally anything but interesting, the last one asked me if I had visited a Kohls or some other department store in the last month (spoilers, no). The 2nd, 3rd and 4th question on the interesting survey are below:
I honestly did a double take as I didn’t expect this to be used to train Google’s ML/AI stacks, though honestly, it makes perfect sense. It’s easy to fire off a survey to a reasonably sized population (as of this writing it has 587,977 reviews on the Google Play Store). On top of that, it costs them as little as .10 cents a survey to be able to help with understanding searcher intent, query disambiguation and potentially other problems that arise when a machine is in control of search.
In fact, as an outcropping of the holocaust debacle, they announced that they would better train the machines to spot offense queries. Looking at the 3rd screenshot, you can see that they are doing just that in a particularly heavy-handed and IMHO irrelevant way. Search queries are rarely offense in and of themselves, context is important for understanding language in the first place. This is why things like structured data are so critical for the future of search as they provide context.
Google’s Love/Hate Relationship With Humans
“The algo is god and that puts the damper on putting real, trained humans on the task.”‘
One of the reasons I love Mike’s work is that he never fails to keep Google in the proper context. They are a billion dollar mega-corporation that is legally driven to provide value to their shareholders and they primarily do this through selling ads. Other products like Nest, Google X etc made less than 1% of total revenue, while ad sales made up ~90%. So when a massive ad network fetishizes everything as a software problem, and has the ability to execute on them, that leads to some really bad outcomes.
A perfect example is Google Photos labeling black people as Gorillas because of their AI/ML stacks. The racist history of comparing black people to primates is a distinctly human problem and one that needs to be addressed at the human level. This is why more and more social science scholars are calling for inclusive teams as a way to combat the bias of applications mainly written by white men. I highly recommend checking out Safiya Umoja Noble book “Algorithms of Oppression” for a deeper dive in the subject matter. Just look at how a former member of the Google Webspam Team views how that team operates:
Honestly, I think this view fails to properly grapple with human nature. Humans are flawed, they do stupid things. When given lots of power these stupid things can have massive repercussions. Edward Snowden made us all acutely aware of how this works with the NSA, where lack of oversight and transparency led to individuals abusing their power and public trust. Google’s is not very transparent, Google does not have 3rd party auditors to make the public feel safer about abuse.
This is IMHO why Google needs to be regulated, but that is a subject for a different post.
So What Does This Mean For SEO
Seriously, don’t trust the machine to get it right. They may sometimes, other times they will fail. When they fail you aren’t going to get any help or support from them because that’s a human solution instead of a programmatic one. Google regularly rolls out changes and updates that affect the livelihoods of hundreds of thousands of people. What do you think their acceptable failure/defect rate is for search changes? 20%? 5%? Either way, people on the edge or who may be wrongly affected have little recourse in order to rectify their plight. And a lot of times, it’s not even an accident. Maybe they intentionally crushed your business, but don’t worry, it isn’t personal. Just gotta keep people on a SERP page and sell more ads.
If your SEO strategy is “Google will get it right most of the time” and you are in that 5%/20% bucket, my condolences.