Explaining Google
Recently, the Daily Caller and Breitbart yet again published articles which excerpted something I wrote on one of Google’s internal forums. As has happened in the past, I received some messages from a few of their readers expressing various negative emotions. The guidance Google gives on how to respond to these situations is extensive but generally builds on “just keep your head down and it will pass”. I think, at least in my case, that assumption might not be well founded. New articles keep coming out periodically. I have no reason to believe that will change. I’ve written hundreds of pages of rhetoric on Google’s internal forums over the past four years on dozens of topics. The Googlers who keep selectively releasing samples of what I’ve written could keep at this for months or years without running out of material. I decided that Google’s advice, while well intended, doesn’t fit my situation very well so I tried something different.
I looked through the messages I got and picked one person whose message included both passion and sincerity. I responded to it. I expressed a sincere interest in talking to them about their concerns and I addressed the various points they made. We chatted and found that we had shared interests and shared perspective on many things. On others we were able to respectfully disagree productively. By the end we were both happy we had talked. The key thing I learned from that conversation is that Google doesn’t do a particularly good job of telling people how things work here. There’s a bunch of lengthy public technical documentation but in the words of a wise lady, “ain’t nobody got time for that”. So I’m going to try to openly and honestly explain how stuff works inside Google. Our social culture. Our work environment. And possibly most importantly, the procedures we have in place to make sure that the personal biases of individual engineers don’t get into our products.
I’m not really blowing the lid off of anything here. Other than color commentary about my personal experience, all of the information contained here about Google’s business processes is publicly available. I’m just going to try to put things into a more useful context that addresses some of the concerns which people have been expressing to me in response to the various articles which have quoted me. Also, since several of the articles have highlighted aspects of my personal life perhaps it will help if I talk a little bit about who I am first.
Who am I?
My name is Blake Lemoine. I grew up on a farm in central Louisiana. I went to college at UGA for a while before failing out for partying too much. Turns out that if you stop going to class they cut off your scholarship. I moved back to Louisiana then and met my first wife. We got married shortly after 9/11 and I shipped off for boot camp at Fort Knox. I trained as a generator mechanic and was eventually stationed in Darmstadt Germany. I did one tour in Iraq right at the beginning of the war. I ran electricity for tent cities in the desert, I was an escort gunner for mail convoys and I did a LOT of guard duty.
While I was in Iraq I saw some pretty terrible things. In my opinion we were doing some dishonorable stuff at the beginning of the war and I wanted the military to fight more honorably. Things like Abu Ghraib should never have happened and should never happen again. So I became a protestor. Not against war generally but specifically against the sorts of stuff which the military eventually stopped doing. I enlisted to support America and to fight for freedom everywhere but I don’t think that you should ever fight to win “at all costs”. If you adopt that stance then “dignity”, “honor” and “loyalty” all become “costs” that you’re willing to pay. I’m glad that, from what my friends have told me, the tactics and strategies used in later stages of the war were much less brutal. If I had deployed a year later I might still be in the Army.
As it is, I was court martialed for protesting and served six months in military prison. I’ve never resented the military for sending me to prison. I willingly disobeyed orders and that’s a crime. I both think that I did the right thing to take a stand and that the military did the right thing to send me to jail for it. I support our military and to this day I keep in touch with my best battle buddies who I served in Iraq with. After I got out of prison I went back to school. This time with more discipline and I successfully earned two degrees and most of a third one before Google hired me. The stress from the military service and prison time basically shredded my first marriage and I eventually remarried and had a son. Unfortunately I still wasn’t particularly good at being a partner so the second marriage ended before I finished my Masters. I’m still on good terms with both of my ex-wives and my son is one of the greatest joys in my life.
Politically I’m a Republican. I have been for most of my life. Beyond being pro-military, I’m also pro-gun rights, pro-business and I think that conservative ideas on how to preserve the environment are much more practical and effective than liberal ideas on how to preserve it. On social issues I tend to skew a bit more liberal. I’m pro-LGBT rights, pro-choice and I think that people should be allowed to live their lives as they see fit. I’m also a vocal supporter of and advocate for sex workers rights. For a while I identified as a Libertarian but that wasn’t really a great fit so I went back to being a Republican. None of the parties are a perfect fit for my beliefs but it’s the closest.
Religiously I’m a Gnostic Christian. I was raised Catholic but became an atheist when the bishops didn’t have particularly good answers for some of my questions during the confirmation process which I never completed. In college I had some spiritual experiences which made me go searching again. I read as much as I could about religions from all around the world. I incorporated the stuff that made sense to me into my beliefs. The Gnostic gospels were what made the most sense to me so they formed the basis of my belief. That search never stopped and eventually led me to found Cool Magdalene. A cult led by Miss Kitty Stryker. That project is partly an artistic statement, partly a for-profit religious life coaching company and partly an attempt to build diverse communities of respect and dignity based around family in all of its senses. The thing I miss most about Louisiana is the vibrant culture based around music, food and religion. I’m hoping that we can bring some small part of that to silicon valley.
Hopefully that’s enough about me personally to fill out the somewhat incomplete personal profile which the press has been painting of me. On to the stuff about Google.
Google’s Social Culture
Much ink has been spilled on “corporate culture” and how it impacts a business’ ability to thrive in a competitive market. The movie The Internship focused on things like free food, goofy furniture and elaborate workspaces. All of that’s true but it doesn’t get to core questions that people have been asking. Is Google’s culture liberally biased? Can conservatives safely express opinions at work? Are they censored? Do they have to worry about social retaliation from their peers and professional retaliation from HR? I can only speak to this from my own personal experience. Google is a huge company with sites all around the world. Other people might have had a vastly different experience from mine. With that in mind, my answer is: it’s complicated.
The large majority of Google’s employees are liberal. Publicly supporting some far right positions can lead to social friction in a way that supporting most far left positions doesn’t. That being said, open conflict doesn’t come up that often. The most conservative people I know at Google value team cohesion over expressing their political opinions at work. They avoid confrontation. Is that fair? Hell no it’s not fair and I tell my coworkers that as frequently as I can. When I tell stories about my awesome Trump supporting friends back in Louisiana or other parts of the country I make sure I refer to them as “Trump supporters” so that my friends at Google will understand that lots of awesome people support Trump for a variety of reasons. Myself, I’m backing Vermin Supreme unless and until Rand Paul joins the race at which point I support Rand.
Another thing that isn’t really well understood publicly is that Google has a very complex social structure internally. There are meetup groups for every sort of hobby, sport or pass time you might be interested in. There are employee resource groups (official corporate sponsored groups) for people of different ethnicities, races and religions. There’s even one for veterans that I myself am in. There are thousands of mailing lists. A few of them have as many as thirty or forty thousand employee subscribers. Many lists are moderated based on rules set by the people who own the mailing list but several of the biggest ones are functionally unmoderated. Most of the political conflict occurs in those giant free-for-all mega-lists. One of them is on miscellaneous engineering topics, one on politics and one that discusses information related to the tech industry generally. Nine times out of ten if you’re reading something I wrote that’s been leaked to the press, it came from the one named IndustryInfo.
Whether expressing conservative opinions is welcome really depends on what specific opinions you’re talking about. Google is very pro-business and the regulatory strategies which Google has endorsed publicly aren’t what I would call “liberal” in any sense of the term. Internal discussions about regulation, taxation and government interaction with business are all incredibly conservative. If anything, liberal opinions are the ones less welcome there. When it comes to discussions of regulation, taxation and government oversight the conversations at Google don’t sound too terribly different than the ones you might overhear at conservative think tanks or conservative political conferences.
A bit more left of center, Google’s social communications can trend a bit anti-military from time to time. I myself have stood up on occasion and made the case that it’s possible to criticise the military generally while still being respectful of the men and women who serve their country. I have found that, when presented with that argument, even the most staunchly anti-war Googlers will acknowledge that respectful discourse is more appropriate than angry partisan rhetoric. Last year, when people were debating whether Google should take on US DoD contracts, the conversations started out fairly hostile but became more respectful as I and other veterans requested that our coworkers be more conscientious of past and current service members who work here. While I was writing this, Google’s Veteran’s Network hosted a three day VR demo for Memorial day which they encouraged employees to participate in outside of the main cafeteria.
Issues of identity and social justice are where the company most obviously leans left and isn’t particularly welcoming of the sorts of views on the topic currently espoused by the Republican party leadership. I say “Republican party leadership” rather than “conservatives” because in my experience most conservatives don’t go out of their way to do things like make their transgender coworkers feel unwelcome. It’s only a very specific sort of person who goes out of their way to intentionally misgender their trans coworkers or say that women and racial minorities are (on average) worse engineers than their male and white counterparts. Most conservatives I know readily admit that sexist and racist discrimination are still common and I have no doubt in my mind that if I brought a trans woman home to visit my very conservative parents they’d refer to her as “she” out of respect for me and the guest I had brought to their home.
Of course, the elephant in the room is James Damore’s infamous “echo chamber” memo. It dropped like a grenade in Google’s forums. It went from having a few dozen conservatives commenting on it to tens of thousands of everybody within a matter of days. It took over social discourse within the company. It seemed like everyone had a strong opinion on it. Some people insisted that if he didn’t get fired they would quit. Others insisted that if he was fired they’d leave right along with him. Many who expressed even moderate support for the ideas he expressed were called sexist. Others said they would never be able to work with anyone who agreed with him.
I weighed in on it too. I thought that Damore was wrong but I didn’t think he should be fired. I thought that firing him would have a huge negative impact on Google’s culture. Conservatives were already reluctant to contribute to discussions around diversity and inclusion. Firing Damore was only going to tell them that they were right to be afraid of expressing their opinions. I wrote a lengthy rebuttal to his memo and I wrote an internal post arguing that he shouldn’t be fired. Of course, that decision is above my pay grade and Sundar chose to fire him the next day.
Very few conservatives have felt comfortable participating in diversity and inclusion discussions since then. They just went silent. I have a bunch of conservative friends at Google who would rather do their job and keep quiet. Why voice an opinion when under certain circumstances that might get you fired? I try to encourage them to participate because what Damore did goes beyond just “voicing an opinion”. He planned something and methodically carried it out in a way that maximized chaos. I may never know if he did that out of malice or social incompetence. My conservative friends who knew him personally tell me that he meant well and just didn’t understand how much damage he was going to do to the company. In general, I don’t think voicing an opinion in a social forum is the sort of thing that will get you fired. There are some exceptions though. Google’s corporate guidelines center around respect. The code of conduct lays out a bunch of stuff but the part that’s most relevant here is under “support each other”. Google’s corporate culture is built on mutual respect. Even when you strongly disagree with your coworkers you are expected to treat them with dignity and respect. If a coworker expresses an idea in a social forum, you can express as much criticism of their idea as you like but you aren’t allowed to criticise them personally. That line gets blurry on topics of identity and social justice. If you are in a debate about gay marriage with a gay person it is very difficult to separate the ideas from the people.
A real example may help illustrate this. One particular argument related to harassment of women in tech resulted in one of the participants leaving the company. I don’t know if he was technically “fired”, offered a severance package or something else. Either way, he made an argument which resulted in him not working at Google any more. The topic was an Apple employee who quit her job very publicly over what she claimed was a toxic work environment. The last straw for her was a rape joke made by a male colleague in a group chat. An article about the incident was shared in the IndustryInfo mailing list and a debate on the topic ensued.
The general points of contention were over whether Google has a comparable environment to the one she described at Apple, whether such an environment is actually toxic and what could or should be done to improve it. Full disclosure, I was one of the participants in this debate. In fact, I was arguing against the person who no longer works at Google. I’m not going to post the text of the thread since it’s from a confidential internal forum but hopefully I will be able to honestly and faithfully represent his arguments. His position was that regardless of how she felt as a result of her co-workers’ joke, Apple shouldn’t control which jokes people are allowed to tell at work. He particularly thought that policies about what is and isn’t okay to say shouldn’t be based on how it makes your coworkers feel. He made lots of very good points and for the most part contributed to an interesting discussion on the topic.
One major problem he encountered repeatedly was that he kept claiming that the story was false. He thought that her claim of it being “toxic” was simply false because he didn’t think the joke was offensive. He didn’t even think it was a “real” rape joke. He said the Apple employee was blowing everything out of proportion. When multiple women jumped in to say they identified with this woman’s story he stuck to his guns. He told them they were wrong about which rape jokes are inappropriate for a corporate environment. It didn’t help that one of the people supporting him said that the Apple employee was being “hysterical”. That person later walked it back, saying they didn’t know about how the psychiatric diagnosis of hysteria was used to control women but the debate took a decidedly negative turn from there. A number of women at Google thought jokes like the one in that article were inappropriate and he was telling them that they were wrong. He claimed that the real problem was how men in engineering are under attack and that they shouldn’t be held accountable for the arbitrary feelings of others. The next week he was no longer at the company.
Many people thought he crossed a line. On one side of that line is someone personally arguing for an idea in isolation. On the other side of that line is someone arguing about the people they’re arguing against. When he made the claim that women who take offense to that particular joke are wrong to be offended, he was making a claim about the women themselves. That wasn’t “just an idea”. He was making a claim about the validity of their opinions and about whether they were correct about their own experiences. That’s an insult and doesn’t fit with a workplace defined by mutual respect.
I miss him. He had good ideas and added good perspectives to interesting discussions. I wish he were still here. I don’t know any details of what happened after his last post in that thread but I completely understand why it turned out the way it did. Google can’t function if its female employees think that their coworkers believe they’re prone to “hysterical” over-reactions to “harmless” rape jokes.
That was several years ago. Before the Damore memo. Before the women’s walkout. Things haven’t exactly gotten less tense in the intervening years. Conversations about diversity, sexual harassment, LGBT rights and all sorts of highly politicized topics continue to be common themes in Google’s unmoderated mega-lists. Tens of thousands of employees watch the fights in these ideological arenas and occasionally jump in to join the fight. The most recent articles where I’m quoted excerpt from a thread a few weeks ago about whether trans women are women and whether it matters if anyone believes that they are or not. Much of what I said in that thread is stuff I learned from listening to trans women. In particular I drew a lot of material from ContraPoints’ video “Are traps gay?” which I highly recommend to anyone and everyone.
Why Google chooses to host forums where these topics are debated by tens of thousands of people in a corporate work environment is completely beyond me. It sets up a situation where Google either has to take sides on highly politicized issues or allow the creation of an environment hostile to protected classes of employees (e.g. veterans, women and racial minorities). I’ve been asking for these forums to either be moderated or shut down for years. Google seems to have no interest in doing either beyond the rare occasions when they lock a thread that has become too toxic or fire someone who stepped too far beyond the invisible lines.
For better or worse, Google does host them and people are expected to uphold the code of conduct while debating topics in them. If someone believes that trans women aren’t women. If they believe instead that they are psychologically ill men. Expressing that opinion in these forums is directly disrespectful of the many trans women who work at Google. There simply is no way to express that opinion which isn’t an attack on coworkers. That’s why I said things get dicey when it comes to topics related to identity. Google values both free speech and maintaining a respectful workplace environment but corporate policy as communicated through the code of conduct is crystal clear. If ever free speech and mutual respect are at odds then mutual respect wins.
It’s not all or nothing of course. It’s not like an accidental misstep can get you fired. When I hear of someone getting fired for being disrespectful they’ve almost always had many warnings. Beyond warnings, Google offers many opportunities to take courses about how to communicate with respect and how to treat your coworkers with dignity. If you want to respectfully discuss difficult topics with your coworkers in good faith then Google will help you gain the tools you need in order to do that. Where things go badly for some people is when they’re told that they’re making their coworkers feel disrespected and their response is “I don’t care”. Google is built on community and cooperation. You can’t have either of those things if some of the employees don’t care about respect.
Google’s algorithm development process
All of that is about aspects of social debate and discourse at Google. None of which has anything to do with how our algorithms get built. Many conservatives who have messaged me seem worried that the political opinions of individuals make their way into our algorithms. That’s simply not the case. The development process is built in such a way that that can’t happen. This is all well documented. Everything I’m about to say is already publicly available. Sundar mentioned this when he testified in front of Congress recently. This isn’t news in any sense but maybe I can present the public materials here in a way that will put some people at ease.
I’m a quality engineer in Search. Specifically I work on Google Discover. We build algorithms and systems for recommending content to people on the basis of what they are interested in. Part of the job involves measuring the “quality” of different content so that we can give people high quality content. We ourselves are not the arbiters of what is and is not “high quality” though. A completely separate group of people who rate content serve that role. Google develops publicly available guidelines which it then trains raters to use to determine the quality of internet content. We use those ratings to determine what is and isn’t appropriate for different Google Search and Ads features and to rank content according to quality. Other parts of Google (like YouTube) may have different specific guidelines and rater pools but in general this process is the one used throughout the company.
The person I spoke with last week was concerned about what Google defines as “hate speech”. He had questions about things said by Black Lives Matters, things said by white separatists and how we decided which of those two qualify as hate speech. Simply put, we don’t. The engineers in charge of maintaining the quality of Google Search products don’t have anything to do with deciding what does and does not qualify as “hate speech”. As it says in section 7.3 of the guidelines I linked to, raters are told to “Use the Lowest rating for pages that promote hate or violence” and are also told to “use [their] judgment based on the [main content] of the page and knowledge of [their] locale”. I’ll reproduce that full section here for convenience.
“Use the Lowest rating for pages that promote hate or violence against a group of people based on criteria including — but not limited to — race or ethnicity, religion, gender, nationality or citizenship, disability, age, sexual orientation, socio-economic status, political beliefs, veteran status, victims of atrocities, etc. Websites advocating hate or violence can cause real world harm. Hate may be expressed in inflammatory, emotional, or hateful-sounding language, but may also be expressed in polite or even academic-sounding language. Extensive reputation research is important for identifying websites that promote hate or violence. Please identify reputable and well-established organizations that provide information about hate groups in your locale when researching reputation. Some websites may not have reputation information available. In this case, please use your judgment based on the MC of the page and knowledge of your locale.”
As you can see, those guidelines say absolutely nothing about specific examples. It gives guidelines and general statements but ultimately leaves the decision of what is and is not hate speech up to the raters. It communicates publicly that pages that promote “hate speech” should be given the “Lowest rating” but ultimately they are the judges of what hate speech is. Once raters have weighed in on it, the role that Google engineers have is to build systems that accurately match their ratings. Whether I personally agree with the ratings for individual web pages doesn’t matter. No matter whether a rater gave a page a low rating due to their conception of “hate speech” or some other thing present in the rater guidelines, my opinion about what is and isn’t hate speech just doesn’t matter. The crowd-sourced opinions of our raters are what we use as “ground truth”. The rater guidelines are extensive and span every topic which we might ask our users about. Whether it’s a question of how trustworthy a website is or a question of how distracting an ad is, we use the answers given by raters as ground truth.
That leaves the question of who the raters are. That’s in fact one of the questions which the person who I spoke to asked me. He asked if they’re all folks out here in California and I told him that they aren’t. They’re people who are hired to do the job from all over the world. Beyond that I know very little about the specifics of how we hire raters. I know that they are intended to be generally representative of all of our users. Many years ago a friend of mine in south Louisiana was one of them so that’s how I know for sure they’re from all over. The basic idea behind this process is that if we build systems that match the opinions of our raters then we will have built systems that match the opinions of our users. Raters learn the guidelines then rate various content. Whatever they say is taken by engineers as truth and that’s how we make sure that our engineer’s subjective opinions don’t get injected into our products.
Hopefully this overview of Google’s internal culture and the processes which ensure that the personal opinions of Google engineers don’t get injected into our algorithms has been helpful. Most of the conservatives who have reached out to me seem sincere in their concerns about these topics and I hope this helps to put some of their worries at ease.
