I’m returning to the open discussion thread format, which I tried out earlier in the year. In that first thread, I raised the topic of death and digital media, taking, as a point of departure, a viral tweet from a student who discovered his online professor was, in fact, dead. A great discussion unfolded, and I thought the experiment was a success. Here, then, is another, long overdue discussion thread for you. Comments are open to anybody who wants to jump in.
This time, we take our point of departure from a somewhat less grave situation:
I hope the context—an American minor league baseball game—doesn’t put too many of you off. What you’re watching is a rather egregious third strike call made by a robot umpire. I’ve been thinking a lot about this clip ever since Will Oremus posted it on twitter with the following observation: “this is a beautiful visual metaphor for the perils of automation, right down to the body language of the batter who can't believe he struck out because of a dumb robot umpire and the human umpire who's like ‘hey i just work here, take it up with the algorithm.’”
Here was my own initial comment: “What strikes me here is that disbelief and resignation replace anger. Anger is pointless, it has no one to attach to. This is just a minor league game, but extrapolating to other spheres of society, the consequences appear demoralizing.”
You don’t need to be an avid baseball fan to know that under ordinary circumstances such call would likely elicit a vehement response from the batter and his manager, with a good chance that one or both might be thrown out of the game. Of course, similar scenes play out in other sports involving judgement calls from umpires or referees.
My interest in what happens here, of course, extends beyond the game of baseball (although I am interested in what this means for the game). I want to use it as a case study for the integration of automation into society more broadly. So here are some possible questions and angles of approach.
Is the emotional dynamic here suggestive of the way automated systems may demoralize human beings in the face of their determinations? What exactly is being outsourced to the machine? Most evidently judgment. But also responsibility? Is this were the anger comes in? Does outsourcing judgment amount to an evasion of responsibility on the part of human beings?
In this case, the machine erred and obviously so. But I’m perhaps more interested in what happens when the machine becomes more accurate on average than human umpires. You can read this article in the New Yorker for some more context, but it appears that this may already be the case. Would it then be reasonable to replace human umpires? This seems to presume that accuracy is the preeminent or even only good at stake. But is it? Are there other goods that may be lost in the pursuit of accuracy?
What would be the media ecological approach? How would we benefit from seeing this as not just the introduction of a discrete technology into an existing context but as something with ecological, that is ecosystem-wide consequences for the game as a whole?
Also, as philosopher Alvin Noë put it in the New Yorker article, “What we’re seeing in baseball is something that is kind of a core dispute in Western civilization. It really is about ‘What is objectivity?’ Is objectivity something that is physical? Is it mathematical? Is it knowable?” Is the kind of judgment involved here susceptible to computationally derived solutions? Does it exclude other functions of the strike zone in relation to the game? Are there cases where the correct solution is not the good one?
Okay, so there you have it. A general field of discussion with a few more specific questions to consider. Again, all are welcome to comment. I look forward to seeing where this goes. FYI: I won’t be chiming in much through the first couple of hours, but will be able to participate more actively a bit later in the afternoon. Cheers,Michael
Are we sure that outsourcing judgments to machines is that novel?
Bureaucracy is a rule-bound system that, in the end, is designed to make decisions based on available choices in one context or another (deliver a service, access a fund, etc.). Are today's bureaucratic institutions any less of a machine than a computer algorithm, which is also rule bound (machine learning is merely a set of techniques for letting machines infer rules and, thereby, enable it to classify future observations).
Representative democracies are methods of outsourcing judgement, too. I neither have the expertise nor the time to, for example, make sound decisions about how best to allocate tax dollars to maintain clean water, pick up trash, hire the right educational administrators, etc. Instead, I vote for the people to whom I outsource those choices.
I'm also not sure that demoralization in the face of evidently unsound decision-making at the hands of a machine is novel, either (see Kafka on bureaucracy :)).
What I find most troublesome is that no one is questioning the authority of the machine's decision. One might argue that the same machine is used to judge balls & strikes for both team – so perhaps it's some sort of fair. But it's also evidently wrong – and not just technically wrong but, because of the authority placed in the machine, morally wrong.
In the case of human refs, there is at least someone with whom to argue and who recognizes you (the manager, say) as someone with whom they're arguing! Baseball has managed to make such arguments a spectacle, but at least those arguing recognize each other as people with whom they're arguing. Machines neither have self-recognition nor recognize others nor, consequently, argue (and change their decisions as a result).
So I'm ok with outsourcing judgement – for society to scale, individuals have to do that. But what's not ok is mindlessly outsourcing authority over decisions and, thereby, any ability to dialog and reason together.
One element of this complex assessment is the power dynamic. With a regular umpire, you might hope that there's some way to negotiate a bad call. The fans boo, the manager talks to him, there could be a video review, etc. With the context around this algorithm, it appears that the player is powerless to appeal an obvious mistake. It is like trying to get justice from a bureaucracy which is impossible to navigate through. The lack of procedural justice is part of why the outcome seems so wrong - though the outcome is also incorrect.
Another side of this: once the system gets better, it may acquire so much trust and power that even on close calls, we might start to revise our assessments to agree with the algorithm. This takes away another sense of agency and power: who am I to trust my eyes and judgment when the computer says otherwise? It has analyzed one hundred billion pitches, how could I know any better?
You could say this is hardly anything newer than the red light camera. But humor me as a longtime soccer fan who has had to debate this around VAR, but the entire concept rests on an implicit and largely false assumption about the existence of a single version of truth. Not to be a moral relativist about it, but is one perspective more "true" than another?
Sports tech rests on this belief that we can coldly and infinitely splice space and time to get a closer and more accurate representation of one singular truth (assuming one exists). This assumption is a lot more philosophical than beer-guzzling sports fans pause to notice.
In soccer VAR at least, there's a belief that more eyes and projected parallel lines bring us closer to a singular truth. But that truth is typically at slow-motion speeds (something that doesn't exist to begin with) and is examined from a non-human panopticon perspective, judged by viewers behind television screens who aren't even physically present as witnesses to the reality of what happened. And, to Alvin Noë's point, is there truly an objective truth free from an observer effect?
Slow-motion replays, for example, can make an illegal contact appear harder and more deliberate when compared to reality at normal speed without an intermediate screen. Is its truth any more valid than what fans in the stadium witnessed on the field? We act as if we believe it is, but it is dubious. Slow-motion replays can still be 50/50 and subject to interpretation, just as the dress can look yellow/white or blue/black depending on the observer.
Hence I've sided against VAR and instead embraced human fallibility as integral to the human experience of the event. Sports are played for humans at the stadium, not for cameras projecting to TVs thousands of miles away from the event.
In a broader sense, automation allows humans to outsource judgement but also ethics and even morality. The roots here really go back to anti-empirical Cartesian logic combined with Aristotelian zero-sum dualism: we wish for life to be a 1 or a 0. But the reality is much more quantum, typically holding multiple states of truth at the same time.
I just ... prefer the game with the umpires. What is the role of that preference? When I think about my favorite baseball stories, many of them have to do with the human drama of the game, whether it's Lloyd McClendon going nuts at an umpire, Randall Simon hitting a mascot, Miami's "grand illusion" hidden ball trick in the CWS, Dock Ellis tripping balls on LSD while he pitches a no-hitter, Jason Giambi sharing his slump-busting gold thong with teammates...umpires and their fallibility are part of that drama. The game is already in trouble for becoming more mechanistic, more inflexibly directed at optimizing for specific pathways (spin rate, launch angle) -- which I've participated in and cheered in certain ways -- but anyway, I can go on about the baseball part, but the baseball part is just an example. What is the role of _preference_ -- because I wonder if we've lost any ability to really actively solicit and follow preference in these matters, many of which are subsumed under auspices of techne or otherwise fenced off from public questioning and democratic rule?
I’m a fiction writer and professor, and I taught a course last semester on “AI in Fiction.” Several my students developed a theory that when AI are “moral actors” we want to consider them people — with all the complexity that comes with that. But if they are disembodied or not human shaped, we don’t know how to yell in their face, so to say; we don’t know how to express our moral outrage. Culpability can’t find a home. So disembodied AIs that make moral decisions become godlike in the sense that they make decisions that can outrage our morality, and we are as powerless to argue with them as we are to argue with a volcano or hurricane.
This is, I think, a very interesting case study in what we as humans want in the sporting arena. Often these technologies are introduced to ensure that sporting justice is done - and indeed, accurate technologies (unlike this one!) would certainly ensure that, to the letter of the law, the correct result is achieved. Yet the fact that so many sporting fans and players find this so dissatisfactory (witness the outpouring of anger in the UK when VAR ruled scorers offside when they had infringed by the length of a big toe) suggests that what we really want is *human* justice to be done. We crave an arena where someone rules with justice and righteousness, even for a couple of hours - and that someone must be a Someone, a person to whom we can appeal, who understands the distinction between mechanical, automated "correct answers" and the more nuanced and contextually subtle concepts of fairness and justice.
I wrote on this a little last year (https://semperlectio.wordpress.com/2020/04/20/the-hallowed-turf/) and found this quote from Thomas Boswell helpful:
“Born to an age where horror has become commonplace, where tragedy has, by its monotonous repetition, become a parody of sorrow, we need to fence off a few parks where humans try to be fair, where skill has some hope of reward, where absurdity has a harder time than usual getting a ticket.”
The trouble with automated justice is that, precisely because it is automated, it is frequently absurd to the human observer or participant. In the rich human context of what is right and fair in a particular sport, conditioned by years of history and shared understand (often inarticulated, even if it were possible to be articulated), a technically "correct" decision might seem frankly ridiculous.
(While I have you - I'm teaching a course on the ethics of technology from a Christian perspective here in the UK, and I've found your insights superbly helpful. Thank you.)
"Are there other goods that may be lost in the pursuit of accuracy?"
This raises the question: What exactly is the purpose of an umpire? Simply to be an accurate judge of the rules? This may seem like the only measure of good officiating in any sport. However, as someone who has officiated football and basketball, I can say that making accurate calls is only part of the job, and perhaps not even the most important part. Veteran officials often speak of "game management" as a critical part of doing the job well, and it is a big part of our evaluation as well. Game management means keeping the emotional tenor of the game in check by ensuring the appearance of fairness. This means that officials call every game differently, depending on the circumstances. Very emotional, physical games must be officiated more strictly, and I don't think fans realize how close these games come to devolving into pure chaos, if not for good game management by the officials. And yes, this also means makeup calls. All good officials make makeup calls because they know when they've made a bad call and they understand that accuracy must be balanced with fairness. But an algorithm doesn't know when it's made a bad call and therefore can't make up for it.
The truth is, accuracy and fairness are two very different things, and we are choosing to maximize one at the expense of the other, simply because accuracy is easier to quantify. It turns out that fairness, or justice, is very difficult to quantify, and any attempts to do so thus far have produced very questionable results. Look no further than the algorithms used to determine court sentencing, loan qualification, etc. I would argue that these algorithms aren't even attempting to quantify fairness. They are simply conflating it with accuracy, something that can be quantified. Of course, humans can make the same errors in judgment, but at least humans can be held accountable.
Unlike the introduction of video review, submitting to the judgment of an automated system as the final arbiter of balls and strikes would fundamentally change the nature of baseball, precisely by undermining the authority structure in which it operates. Even when video review leads to calls being reversed on the field, the decision ultimately depends on the considered judgment of the umpire(s). In other words, video review leads to a kind of enhancement of judgment. With the automated strike zone, however, this judgement is entirely superceded by an algorithm, which replaces, rather than enhances judgment.
One important effect of this on the game will be a fundamental change in the meaning of "ball" and "strike." Although at first the algorithm will be adjusted to track, as best it can, what we now understand the strike zone to be, eventually the strike zone will be defined by whatever the algorithm determines. Which is more or less to say that the terms "ball" and "strike" will become meaningless, since they no longer correspond to any kind of human judgment about the nature of things.
The automated umpire is being used within a minor league game. Though the camera doesn't pan around enough to provide a glimpse into what the audience looks like in that clip, from what little had been filmed, doesn't seem like there were many audience members around. I don't know much about baseball, but audience seems to be an important differentiator between a major league game versus a minor league game. When anger turned to resignation, is resignation something humans bond over? Is it a 'sticky' emotion? Perhaps part of attending a game is to experience emotional ups and downs with another friend, a group of friends, or a crowd of fans. In a way, we sort of enjoy games because of these elements of surprise, the subjectivity of it all.
"Are there other goods that may be lost in the pursuit of accuracy?";
We could play the game with robot players that could way outperform humans. What would the fun be in that. I am OK with the very limited mistakes that human umpires make in MLB. Also, it should be noted that younger umps call a much better zone than older ones. It has more to do with how they were trained (on a consistent vs. personal zone) and not eyesight. The problem of bad strike zones will continue to diminish as older umpires retire and newer umps with better zones come up. The need for the robots is continuously decreasing.
"Does outsourcing judgment amount to an evasion of responsibility on the part of human beings?"
Getting into the weeds of automation is still new for me, but speaking from my current context (car sales), this is normal. For example, we do *some* of our service on site, but my managers send the vehicles out to a 3rd party site for an actual inspection. Reason? They say it's because of our workload, but it really boils down to them being able to go, "Well, so-and-so inspected it, and it passed," if there's a problem. So all of that to say: I wouldn't be surprised if the aforementioned question ends up being answered positively.
Here's an interesting follow up sent to me by a reader. It's a twitter thread making the case that the pitch might, in fact, have been a strike. Take a look: https://twitter.com/RPimpsner/status/1413589635880660994?s=20. I still think the catcher's reaction suggests that it wasn't, which means, of course, that the played strike zone is not coterminous with the technically defined strike zone.
This whole debate put me in mind of an Atlantic article from a few years ago: https://www.theatlantic.com/ideas/archive/2019/10/how-baseball-teaches-americans-about-law/600454/#anchor1b
The basic thesis of the article is that baseball, like many other games, is a school for life and for a particular kind of life. In particular, the author claims that "Baseball is the most legalistic sport in the world, and its actors—players and umpires—mirror the way citizens and judges interact in the real world of the law. By playing or watching baseball, Americans absorb an abiding respect for judges, courts, and the rule of law."
Might one of the implications of this technology be a change in the way we regard law? In the AI-umpire framework, law is not something in which we participate or which is properly rational -- one does not "reason with" an AI, one either commands it or submits to it. There is no participatory framework in which an AI umpire can participate.
I appreciated the other comments about how a human umpire, aware of his limitations, can try to compensate for his close calls in favor of one team by deciding that the next close call will favor the other team, and that this plays out as a result of the the interaction of fans, managers, players, and officials. This seems like a good school for a democracy governed by laws -- the people may petition for a redress of grievances, and yet there is a law which they acknowledge (e.g. the rule-defined strike zone).
None of us know what it is like to not be part of the dominant species on the planet. This evolutionary step we are in will give us a taste of that humble pie. And it is not a knock on humans, some of my best friends are human, but it is not going to be pretty.
Yes. Once actual machines are more accurate than human umpires they should replace them.
However, what if robots were better batters too?
Ultimately, the question is what would unlimited energy and limitlessly "productive" robots liberate us to do?
I thought that was the point of Marxism? The elimination of drudgery!
Noë's questions are good ones, and they point to why there won't be such quiet submission to the authority of the machines as arbiters of an "objective reality." Baseball is just not that serious compared to something like COVID-19 or American Exceptionalism baked into real imperial armies, navies, and nuclear arsenals on a dying planet. In those contexts, the consensus reality is eventually asserted by body counts and sufficient pain to end all debate and fundamentally change millions of minds. Crude ballistics still trump rhetorical persuasion and algorithms, and behind it all is simply the rusty wire that holds the cork that keeps the anger in.
There are no bigger believers in an "objective reality" than people who believe they possess it and are deeply invested in using it to control others. Beyond relationships gone bad and the garden variety fundamentalist-dogmatist personality, internal to the ailing nation-states of the west we now have the entrenched divisions of the culture wars and politics that allow demonization of others to the point that a kind of civil war and slow-motion collapse is truly in effect, just without the traditional secession or mass revolt. A great bloodletting isn't necessary for catharsis and reconstitution of a shared social reality, but given past performance, western civilization prefers to do it that way. No robot can make it happen; humans will have to. Someone pushes a button or pulls a trigger.
It is really an ecumenical narcissism, this realism that tends toward the division of everyone into John Milton's "church of one." I suppose it is just the late acceleration of what has been going on since before his time. But it also has the effect of making everyone "wrong" if the "correct," good, or truly pragmatic view is to uphold a manifold of divergent but somehow collectively sustained social realities within sufficiently powerful narratives.
The Zugzwang of our predicament just gets deeper as we now have people insisting that willingness to live perhaps permanently within psychologically and socially destructive isolation is what "science" and their truth demands — and that to disagree is sociopathic. If you agree, you engage in destruction; if you disagree, you foment more — even if you don't wish to do harm and are simply living within different values and actuarial calculations with a different conclusion about "greatest harm." Still, you will be lumped with radical libertarians and people who have acutely paranoid conspiracy theories if you express non-orthodox views even while complying pragmatically with the frequently pointless genuflection of masking and second or third shots. Of course in many locales, the orthodoxy and hostilities are reversed.
As the virus turns out to be the new malaria of the north, we may have two camps of people dividing over how to live with it — one in an antisocial disembodied life not worth living in the views of others who often take foolhardy risks and behave irresponsibly. No centre will remain. Many self-lacerating absurdities will follow like employees fired and losing medical insurance for resisting vaccination despite also being required to work remotely from isolation. We have probably all heard the view expressed that the unorthodox should be hurried to their demise by refusing all care as the price of them refusing some care. It is like the practice of punishing students by temporarily exiling them from school or making prisoners' lives harder, or immiserating poor people for being poor. It is what one does — in Hell.
This insanely self-destructive logic also makes a florid appearance in the now universal belief that judges are and must be partial and partisan — or how all sides taken on the most vexed issues are toxic and compound each other for dividends of destruction. In Texas, the concoction of a law corrosive to the legal system that works like a curse — mocking the whole idea of the law and a judiciary — delegates its enforcement, like gun ownership, to any irritable bowel. A culture gone this far is full of these unfixable dirty logic bombs radiating lethal dilemmas.
But this is why I don't worry too much about quiet acquiescence to the machines, or at least not western ones. The death spiral is already here for the Abendlands. It's written into their founding stories and true names. Much of the west is in a Mexican standoff with itself — its own people and principles so why not expect it will all end like an Elizabethan revenge play? It seems to want to. That is the script when you run out of time and the plot comes to an abrupt end. In some sense, the shooting started a few generations ago in the US. The bodies have been falling steadily. It's sheer psychosis, spin, pride, and misplaced power that lets anyone believe this is a country in control of itself — or anything! — and that it deserves control. A sudden geopolitical shift seems overdetermined, maybe on par with the Russian Empire in 1905.
But, as Yogi Berra said, "Sometimes in baseball, you don't know nothing." We can hope to be surprised. But I do not expect to be.
Baseball is broken. The commissioner is trying to change the fabric of the game. The owners are lining their pockets at the expense of player development, honest wages and the growth of the game. The players union only represents the game's superstars, while the middle class and minor leaguers are overlooked. Everything is just sad. Not only in baseball, but in every avenue of life do the plights of freedom and justice seem hopeless. There is little philosophy among the youth and practically no phronesis. It's enough to make one mad with wrath. I don't know how you all stay sane.
I don't know how to be happy. I want to blame society's ills for my personal shortcomings.
I don't really have any helpful observations or anything interesting to say. Jesus Christ is King of Kings. That's the only solace I can find when thinking about cultural criticism.
"Is the emotional dynamic here suggestive of the way automated systems may demoralize human beings in the face of their determinations?"
I feel like this is a good place to slot into the discussion a recent Slate piece on whether or not we might actually need our robots to be jerks now and then.
Just an attempt at clarification:
We ask about the objectivity of either of three categories of thing, roughly speaking: (1) things that have some sort of ability to get things right and wrong, (2) intentional things (generated by the first things) with representational purport, and (3) things that those first and second things can be right or wrong about.
Examples of things in the first category are a human ump and a robo ump; we might ask, "Is that human ump objective, i.e., free of bias, etc.? What about that robo ump?"
An example of 2 is the claim that a pitch was a strike; we might ask, "Is that claim objective, i.e., true?"
An example of 3 is the placement of a pitch; we might ask, "Was that objectively, i.e., actually, a strike, or did the catcher fool the ump with his framing?"
"Objectivity," then comes to slightly different things in these three cases. A definition of what category-1 objectivity is might differ from a definition of what category-3 objectivity is — the former will invoke the notion of a lack of bias, perhaps, which makes little sense to do when talking about category-3 objectivity.
But, of course, even though these modes of objectivity can be distinguished, they can be co-present and connected. For example, we might think that whether a category-2 thing is objective can depend on whether the category-1 thing that generates it is objective.
I hope this helps in getting clear on some of the questions we might ask ourselves about objectivity.
The demoralized response from the hitter may come from having years of experience arguing with human umps. Maybe if he had come up with robot umps, he'd have a different response. Transitions are hard.
Not entirely on topic, but what bothered me most about this video is the hilariously oblivious commentator who went on about nothing for thirty seconds before noticing that ball two had in fact been called strike three. There's a metaphor in there about our woefully impoverished Attention Society.