Home > Uncategorized > Donald Trump is like a biased machine learning algorithm

Donald Trump is like a biased machine learning algorithm

August 11, 2016

Bear with me while I explain.

A quick observation: Donald Trump is not like normal people. In particular, he doesn’t have any principles to speak of, that might guide him. No moral compass.

That doesn’t mean he doesn’t have a method. He does, but it’s local rather than global.

Instead of following some hidden but stable agenda, I would suggest Trump’s goal is simply to “not be boring” at Trump rallies. He wants to entertain, and to be the focus of attention at all times. He’s said as much, and it’s consistent with what we know about him. A born salesman.

What that translates to is a constant iterative process whereby he experiments with pushing the conversation this way or that, and he sees how the crowd responds. If they like it, he goes there. If they don’t respond, he never goes there again, because he doesn’t want to be boring. If they respond by getting agitated, that’s a lot better than being bored. That’s how he learns.

A few consequences. First, he’s got biased training data, because the people at his rallies are a particular type of weirdo. That’s one reason he consistently ends up saying things that totally fly within his training set – people at rallies – but rub the rest of the world the wrong way.

Next, because he doesn’t have any actual beliefs, his policy ideas are by construction vague. When he’s forced to say more, he makes them benefit himself, naturally, because he’s also selfish. He’s also entirely willing to switch sides on an issue if the crowd at his rallies seem to enjoy that.

In that sense he’s perfectly objective, as in morally neutral. He just follows the numbers. He could be replaced by a robot that acts on a machine learning algorithm with a bad definition of success – or in his case, a penalty for boringness – and with extremely biased data.

The reason I bring this up: first of all, it’s a great way of understanding how machine learning algorithms can give us stuff we absolutely don’t want, even though they fundamentally lack prior agendas. Happens all the time, in ways similar to the Donald.

Second, some people actually think there will soon be algorithms that control us, operating “through sound decisions of pure rationality” and that we will no longer have use for politicians at all.

And look, I can understand why people are sick of politicians, and would love them to be replaced with rational decision-making robots. But that scenario means one of three things:

  1. Controlling robots simply get trained by the people’s will and do whatever people want at the moment. Maybe that looks like people voting with their phones or via the chips in their heads. This is akin to direct democracy, and the problems are varied – I was in Occupy after all – but in particular mean that people are constantly weighing in on things they don’t actually understand. That leaves them vulnerable to misinformation and propaganda.
  2. Controlling robots ignore people’s will and just follow their inner agendas. Then the question becomes, who sets that agenda? And how does it change as the world and as culture changes? Imagine if we were controlled by someone from 1000 years ago with the social mores from that time. Someone’s gonna be in charge of “fixing” things.
  3. Finally, it’s possible that the controlling robot would act within a political framework to be somewhat but not completely influenced by a democratic process. Something like our current president. But then getting a robot in charge would be a lot like voting for a president. Some people would agree with it, some wouldn’t. Maybe every four years we’d have another vote, and the candidates would be both people and robots, and sometimes a robot would win, sometimes a person. I’m not saying it’s impossible, but it’s not utopian. There’s no such thing as pure rationality in politics, it’s much more about picking sides and appealing to some people’s desires while ignoring others.
Categories: Uncategorized
  1. Abe Kohen
    August 11, 2016 at 8:06 am

    Can’t wait for your analysis of the short-circuited one.


  2. August 11, 2016 at 9:18 am

    I have had similar thoughts, but I think it can and should be pushed further.

    Trump also gets to hear feedback on how his rallies played out from outsiders (all sides of the press report on it, commenters from all sides give feedback on those reports). Trump only gets broad feedback at the rallies, where he essentially gets to hear the whole crowd and assess what “works”. I haven’t heard that he walks the crowd and talks with individuals. So where does he get his new ideas? If he is doing online learning, where does he get the new data? I am thinking in particular of the controversy with the Khans. The first time he talked about it he said: “Lots of people are saying that [it is curious the wife was not speaking at the DNC]”. I think he is sincere there. Lots of people were saying this indeed, but somehow I had myself not heard of it and he didn’t test that line in a rally.

    Trump gets assailed with messages from all kinds of sources on his Twitter feed. Could he be using that as input of his campaign speeches? Or is he using his tried-and-true technique of 50 printouts from the press talking about him, but with a much expanded scope (Breitbart etc) and different weights from the mainstream?

    Quite possibly the simplest technique he could himself be using to gather useful data is to look at the number of retweets on his idiosyncratic retweet technique, combined with Twitter analytics (“your message got retweeted this much in this state”). His technique is to retweet by turning the content into a tweet he himself owns, which means he gets to mine the analytics based on that content. That would actually give a perfectly logical explanation to his idiosyncratic technique. Makes sense?


    • August 11, 2016 at 9:27 am

      Yes, I should have mentioned Twitter as a secondary training data source.


  3. Michael
    August 11, 2016 at 10:18 am

    The best explanation I’ve seen yet – a machine learning algorithm.


  4. allenknutson
    August 11, 2016 at 10:23 am

    I had thought about this a while ago, imagining how things might have gone if the initial conditions had been different. Imagine if he’d had Bernie’s audience for a while, and had tailored his message to attack from the left in all cases. It’s not a very realistic scenario; I don’t think he could have carried off a very convincing campaign of hope rather than hate, especially with the recent analyses indicating that all his hopeful tweets come from the campaign whereas the hateful tweets come from him.

    As for training data, his staff says he just watches TV all day.


      • allenknutson
        August 11, 2016 at 4:03 pm

        Thanks. I hadn’t realized Gunnar was writing articles like this.


      • Guest2
        August 12, 2016 at 9:57 pm

        Yes, but what does the background look like? This focuses on the foreground, erasing the background. It is incomplete analysis.


    • August 12, 2016 at 3:52 pm

      For decades, that’s what he did. Then, once he wasn’t getting enough attention from Democrats, he switched to Republicans. You want to know what The Donald sounds like when he’s getting input from Democrats? Read interviews he gave ten years ago. His input was from upper-class, wealthy New York Democrats, and that’s what he sounds like. Hates George W Bush and Cheney, thinks Scooter Libby was a fall guy, generally pro-Hillary Clinton, pro-Obama.

      The same sorts of generalized Democratic talking points that we more-or-less liberal people were repeating to each other.

      Liked by 1 person

  5. August 11, 2016 at 10:31 am

    Question: What is rational? Who defines rational? Does the definition ever change? Is it possible for two completely polar opposite solutions to a similar problem to both be rational?



  6. August 11, 2016 at 10:59 am

    @JamesNT, I think “rational” is an adjective which describes a characteristic of a process, in the sense that it either calculates consistently or acts strictly in accordance with a set of given assumptions. As to your question “Is it possible for two completely polar opposite solutions to a similar problem to both be rational?”, sure, in my opinion, if there are no constraints in the given set of assumptions on the manner of solution. On the other hand, it is possible to prescribe goals and enforce them so severely that the manner in which they are achieved is itself strongly constrained. Accordingly, there are several instances in invertebrate biology where similar evolutionary strategies and life tactics have been adopted by unrelated clades, because they work. Note that evolution may not progress using steps which are judged “rational” by an observer, notably a human, but the objectives can still be achieved. It does raise the question and objection to my own definition on whether or not stochastic optimizers can be “rational”.


    • August 11, 2016 at 10:15 pm

      Or it could be that the definition of “rational” changes with our moods and what we want to do at that time. Also, situations and times change so what is rational, at that time, is different from what was rational before. Today, we have discussion on assisted suicide. Didn’t have that 50 years ago.

      What fascinates me right now is Cathy calling almost half the United States population “a particular type of weirdo” despite my previous comments as to why so many support Trump. I’m not saying O’Neil is wrong, at least not entirely, but there is so much more than what is seen on the surface.



      • August 11, 2016 at 11:21 pm

        Anecdotally, I’ve been struck by how many “blue tribe” people who actually conduct the ethnographic experiment of going to a Trump rally find the people there not particularly weird (with the possible exception of the anti-Trump protesters). Random example from a low-effort google: http://www.businessinsider.com/gselevator-trump-rally-nothing-expected-2016-6/#the-art-of-the-deal-29

        (And yes, I’d say that GSElevator is pretty squarely in the cultural blue tribe, rather than a closet Trumpisto.)


        • August 12, 2016 at 9:56 am

          Trump was here in my home town of Wilmington, NC yesterday speaking at the college I graduated from, UNCW. I also heard the same thing from several – nothing like what they saw on the news. Just good ole folks who are patriotic and want their jobs back.

          Of course, we are all aware that the news is the true bias, right? I would take this post by Mathbabe more seriously if she had attended several rallies in person. In person is the only source to trust with all the rhetoric going on. I would not trust any third party sources because they all pretty much have an agenda.


          Liked by 2 people

        • August 12, 2016 at 10:01 am

          Still waiting for Cathy’s analysis of the short-circuited candidate.


  7. mathematrucker
    August 11, 2016 at 12:27 pm

    Trump also appears to be skillful at applying information he receives from seasoned politicians and entertainers. Two that I know he sought and received counsel from before entering the race are Newt Gingrich and Howard Stern (a master of “not boring”). He likely consulted others too.


  8. August 11, 2016 at 4:43 pm

    Reblogged this on Matthews' Blog and commented:
    Party and business interests guide most publications. This view and thoughts have their focus.


  9. August 11, 2016 at 6:04 pm

    Reblogged this on Managementpublic.


  10. August 11, 2016 at 11:01 pm

    Great article! I love this explanation of what makes The Donald tick, and I greatly appreciate how it doesn’t rely on armchair mental illness diagnoses. Now I’m curious to learn more about machine learning algorithms, and what people expect of them!


  11. Klondike Jack
    August 12, 2016 at 1:51 am

    I think an examination of the way the Trump algorithm produces a feedback loop that also affects and modifies those doing the data training is in order. And sorry but yes, I couldn’t help but think of an electric guitar placed in front of it’s speaker as an analogy. Anyway, It seems possible that the data training/feedback loop may possibly be slowly growing in size or magnitude as the message is refined by the separate data training stream from those republican politicians and donors who support Trump and want him to engage in behavior more likely to result in winning the election. They may be considered to function as an occasionally functional governor on the engine, one which keeps its off balance oscillations within safe limits for the overall machine’s short term survival, at least till the day after the election. Also, as Trump acquires the remainder of the original core of the “fringe” elements of his base of support and then begins to rely more on the support of mainstream republicans, the proportion of less “fringey” supporters increases. Perhaps that modification of the data training stream is what may drive Trump in the direction of something resembling sanity which has the potential to further modify the data training as mathbabe described. A caveate: never forget that in politics as in war, the truth is the first victim. My bet is that Trump was expecting to operate like this going into the race. Also to be factored in is the extent to which Trump is functioning as a distraction from and a front man for the republican platform which is virtually no different than what has been desired by the most extreme elements to survive the GOPs infighting and attrition in recent years. All this being said, the basic premise of the article is quite a sound lense through which to view Trumps behavior and possibly his strategy, and it’s also interesting until you realize how scary the dynamic is.


  12. John Robb
    August 12, 2016 at 6:50 am

    Tremendous analysis Cathy. One twist. This approach makes him the perfect candidate for an open source political movement.


  13. Well...
    August 13, 2016 at 12:30 am

    Interesting and (maybe) novel analysis, but seems like it could be applied to many politicians to varying degrees–at least any politician who has ever pivoted or “evolved” on an issue for political reasons. That covers Clinton and Obama, BTW.

    Cathy, do you know any Trump supporters personally?


  14. G.
    August 24, 2016 at 10:33 am

    What an interesting post statement here, from the author and promoter of “How Big Data Increases Inequality and Threatens Democracy”, from which one would naturally infer that she wants to protect democracy.

    Here we have: “A few consequences. First, he’s got biased training data, because the people at his rallies are a particular type of weirdo. That’s one reason he consistently ends up saying things that totally fly within his training set – people at rallies – but rub the rest of the world the wrong way.”

    Well, it seems the author is against Donald Trump for purely following the will of his supporter base, which is supposedly what democracy should in fact be — the hackneyed phrase of “rule of the people, by the people, for the people”. It seems that the author considers Trump’s base who by the way were numerous enough to catapult him into a major party candidacy — all weirdos, “a particular type” to be more precise. It would seem that the author is opposed to a candidacy that is the will of the people, ie, opposing democracy itself. Then why worry about protecting democracy?

    Btw, why so much harping on about democracy? Let us take a look at the text of the US constitution, another much vaunted document: http://www.archives.gov/exhibits/charters/constitution_transcript.html Well, well, not a single mention of democracy, so easy to check now because of “data” techniques, actually just a test search. Maybe the founders did not have democracy in mind?


  1. August 11, 2016 at 1:16 pm
  2. August 11, 2016 at 7:17 pm
  3. August 12, 2016 at 7:00 am
  4. August 12, 2016 at 3:35 pm
  5. August 13, 2016 at 9:47 pm
  6. August 15, 2016 at 10:19 am
  7. August 15, 2016 at 10:20 am
Comments are closed.
%d bloggers like this: