A blog about problems in the field of psychology and attempts to fix them.

Tuesday, May 14, 2013

Psychological Realism and Poker

During my post-doctoral years I played poker very seriously. For a while, my poker library grew much faster than my psychology library, and I became a profitable mid-level player. I have played very little since my post-doc ended, but I think the experience was valuable. For one thing, the mathematics of poker is fascinating, and I still nurture a hope of one day teaching a "Statistics of Poker" seminar. For another thing, I think poker provides an excellent context for thinking through theories of psychology. On the surface, poker seems like it is a game about cards, and on the surface it is. However, you don't need to get much below surface-level to see that poker is primarily a game about the behavior of the other players. The player on your right just put in a big bet: Does he have a big hand? Is he bluffing? What does he think you have, and how does he think you will respond? Given what he might have and what he thinks you might have, if you put in a huge re-reraise, how will he respond? The layers of analysis that can be applied to these situations is fun, but not really on topic for this post (though I talked a little about it here). Instead, I want to delve into a very typical poker situation from the point of view of a psychological realist vs. a dualist.

Note that this is a preliminary analysis that I hope to develop further, and I would love any feedback. 


A Single Hand

To keep the situation simple, we will be watching the most popular poker game of the moment, No-Limit Texas Hold'um. No-Limit Hold'um went from obscurity to ubiquity when some reporters realized that it made for great television, especially when the audience at home can see the players hidden cards. In this game each player gets two hidden cards, and there are five cards placed face up on the table (through a series of reveals). Each player makes the best five-card hand she can out of the seven cards available to her. For now, we will fast forward to the end of a hand between two players, and use an example where the suit of the cards don't matter.

Player 1's face-down cards: 7, 8
Player 2's face-down cards: Jack, Queen
Face-up cards for either to use: 7, 6, 5, Ace, King

Here, at the end, Player 1 has the best hand, a pair of 7s. Player 2 has the worse hand, with a high card (i.e., nothing).

However, the hand started out with a full eight-person table, and Player 2 has been betting aggressively. Thus it might seem likely, to an live observer, that Player 2 either had a good pair to begin with, or that he connected to one of the last two cards, and now has the best hand. Thus, there is room for both players to make some interesting moves.

Now we need some basic terminology:
A bluff is a bet with the worst hand. Poker players intentionally bluff when attempting to trick an opponent into folding the best hand. (There are other game-theoretic reasons to bluff, but we'll let those go for now.)

A value bet is a type of bet with the best hand. Poker players value bet in hopes their opponents will call the bet and give them more money.

In this, the final round of betting, Player 1 leads out with a large bet. For simplicity, let us say there is $100 in the middle already, and that Player 1 bets $50.

Let us look at this situation from the two major approaches to psychology 1) a dualistic, information-processing position and 2) a psychological realist position.

Dualistic, Information Processing
For the players, in the moment, poker is certainly a game of limited information, and so the information-processing model can take us pretty far. Neither player knows the other players face-down cards, but each can guess based on the other player's behavior. Thus, while neither player knows exactly where they stand, they have probabilistic cues from the patterns of betting within this hand, from the patterns of betting in past hands, and from any "tells" they can pick up by observing their opponent carefully. 

In this case, Player 1 is in the same position as a live observer in the room. She is forced to guess at what Player 2 has, and an evaluation of the available "information" leads her to believe she is likely to be beaten. However, she bet anyway. If you asked her why she bet, she would tell you that she was bluffing, and hoping that Player 2 will fold his hand and give her the $100 pot. From a dualistic perspective this is all fine and good. We can never know exactly what Player 1 is doing, but Player 1 knows perfectly (literally, perfectly), and if she says she is bluffing then by God that is what she is doing. Another way to put this is that from a dualistic perspective bluffing is a "state of mind", and she is bluffing. If we wanted to increase our confidence that her verbal report was accurate, at best we might try some careful brain scanning, because the truth is to be found somewhere inside her head.

Thus, from this point of view, we think that it is Player 2's job to try to "get inside the head" of Player 1, so that he can determine that she is bluffing. However, in this case Player 2 knows that he has nothing but a high card, and that therefore he is probably beaten. If he simply calls the $50 bet, there is little doubt that his is handing over free money. His choices are either to fold and give up the money already in the pot, or to try to re-raise the bet and bluff Player 1. The correct decision, the one with the best long-term expected value, can be determined by a bunch of math that we don't need to worry about here. The point is that from a dualistic point of view Player 2 cannot possibly know what Player 1 is thinking, as even the best-case scenario is considered analogous to the game of limited information he is stuck in by happenstance.

Psychological Realism
The major claim of psychological realism is that mental states (processes, events, or whatever you might want to call them) are observable features of the world. That is: Given enough knowledge of the world -- knowledge of what is happening now and/or of the person's behavior across other situations -- you can know what they are doing. For certain. The most provocative way of saying this is that "You can see their minds." A less provocative way of saying it would be to list contrasting mental traits, i.e., "You can know if they are remembering vs. imagining, paying attention vs. daydreaming, intending to succeed vs. intending to fail, trying hard vs. going through the motions, value-betting vs. bluffing, etc." The analysis of the above situation from the perspective of a psychological realist has, I would argue, certain advantages.

Poker, if all the cards were face up would be boring for most players, but it turns out to be great fun for a viewing audience at home. The main technological innovation that lead to the "poker revolution" about 15 years ago was the "hole camera", which allows the viewing audience to see the players face down cards. This allows you to see how the best players play, but more importantly, it allows the viewer to get outside of the limited-information game that the live players are stuck in. This highlights the "psychological" game that is being played.

A person at home, who can see all the cards, knows things that the players do not. For example, this person knows that Player 1 is not bluffing. From a realist perspective, bluffing is a doing-under-a-particular-circumstance, and as Player 1 does not have the worst hand, Player 1 cannot possibly be bluffing. We might speculate as to whether or not Player 1 is "trying" to bluff, but in no case is she bluffing now. (To engage in the speculation we would compare Player 1's behavior in this circumstance to her behavior in other situations that were, in fact, bluffs). In this situation, no matter what Player 1 was trying to do, she was, in fact, value betting. As such, assuming this play is part of a larger pattern designed to take Player 2's money, then Player 1 wants Player 2 to call.This is true no matter what she "thinks" she wants, i.e., what she would verbally report she wants.

Wait, Why Does This Matter?
It matters because the realist approach provides a language that allows you to talk about many important phenomena than are difficult to talk about in dualism-land (also the realist approach is correct, but that isn't the big issue right now). In this case, it allows us to talk about how Player 1 was wrong in thinking that she was bluffing, when in fact she was not. A third party observer, able to see the face-down cards would have accurately reported that she was not bluffing.

Leaving the poker context for a minute, this is the same realist perspective that might allow a therapist to tell a client that they care about someone the patient claims not to care about. The therapist, as an observant third party, can potentially see the patterns in the client's behavior that the client does not see, or does not want to admit to. Returning to poker it is the same perspective that allows us to tell the difference between the poker player who is striving to make the most money, versus the player trying to win the most hands, versus the player trying to avoid being embarrassed, etc.

Poker provides a good contrast between 1) our understanding of other people's behavior based on the situation we might be stuck in at the moment, and 2) the understanding of other people's behavior we could have with a wider view of the world. In live poker, this difference is palpable every minute: The reality of the situation could be revealed at any moment by a gust of wind, sloppy handling, or anything else that turned the players' cards face up. The difference is even more obvious now that televised poker allows viewers at home to see what is going on while keeping the players in the dark.

Crucially, the person watching on TV, who can see the face-down cards, knows what is happening better than the players themselves. Player 1 might think she is performing a bluff, but the people watching TV know that she is not. Player 1 might think that Player 2's long pause before folding indicates that he considered calling the bet, but the people watching TV know that it is just a deceptive show to hide weakness. The dualistic model assumes that we are inherently, by the nature of the universe, stuck in the position of the players at the table; we cannot escape playing the guessing game with regards to the state of world, and particularly the state other people's minds. The realist model assumes that we could, if were in the right situation, directly perceive the state of the world, including knowing what other people are doing, in the broader, mental sense. Because the other person might not be observing carefully enough, or might not be sensitive to the right variables, or might not have access to the necessary point of view, it is in principle possible for us to know what another person is doing better than they know themselves. That is, the realist perspective assumes that, in principle, enough knowledge of the world could always make us like the viewing audience at home.

While this is just a quick sketch, it shows the potential for the poker table to serve a s a context for such analyses. What do you think?




13 comments:

  1. As always, a very interesting read, thank you. Perspective is indeed an interesting subject. A poker game holds a wide variety of them and is a valuable context to discuss perspective in.

    However, I do believe that an objective perspective is not always most accurate and does not always supercede the subjective perspective(s).

    "Crucially, the person watching on TV, who can see the face-down cards, knows what is happening better than the players themselves. Player 1 might think she is performing a bluff, but the people watching TV know that she is not."
    Here, an objective perspective is favoured, but, to P1's subjective perspective, she only has her situationally available information to rely on. Is it not accurate then to describe P1's behaviour based on the reasoning P1 would have given as she is deliberating on what to do before she bets the 50USD? In this specific sense I am not sure I believe we can know what another person is doing better than themselves.

    "Player 1 might think that Player 2's long pause before folding indicates that he considered calling the bet, but the people watching TV know that it is just a deceptive show to hide weakness."
    I have experienced both, knowing that I do not have anything but trying to deceive, and, having something deliberating whether I would bluff or fold. The assumption in the quotation is that P1 is wrong in her perception of P2's behaviour, and this will gain P2 in future hands. If P1 is correct however (and the audience wrong), it will gain her in future hands. Pragmatically then, an objective perspective does not hold any consequences for behaviour, only the subjective perspectives of the situated agents.

    In poker, the perspective of the audience is indeed intriguing, and I believe it to be so because we can see players over-evaluating their own hand, under-evaluating their own hand, over-evaluating others' hands, under-evaluating others' hands, misreading behavioural cues, or accurately perceiving them, based off of the subjective perspective each player has in that situation -which, to make it even more exciting- is carried forward into the many upcoming hands. Successful deception being the very most exciting because it may involve terrible loss, and an "unjust" one at that!

    ReplyDelete
    Replies
    1. I guess it comes down to that, while I do believe we can have direct perception and that we can know what another person is doing better than they know themselves, I am not sure if contrasting between, or exemplifying through, subjective and objective perspectives will lend itself to this. Is what there exists more accurately described from the agent herself or someone observing her? My answer to this is most often that an objective perspective almost always reveals many more options to describe behaviour from than a subjective one -but it doesn't mean that it is necessarily more accurate.

      I believe "Because the other person might not be observing carefully enough... ...it is in principle possible for us to know what another person is doing better than they know themselves." is true, but, an objective perspective gives as much, or little, indication of what is really going on. This is able to be elucidated by contextualising through poker, by non-hierarchically taking all perspectives into consideration. What information is available to each perspective? Which resources are available to each perspective? Which consequences may the available information and resources have on each perspective's behaviour? What is then the most probable description of someone's observed behaviour? In this way, the subjective perspective is taken into account and eventually provides for an objective perspective, rendering the distinction between them fuzzy. After all, the assumption that P2's long pause is X is only one perspective of many that one can take, the information available for the tv-viewer is then biased by the parts of which that is not available for P1 or P2 and may mislead the tv-viewers conclusions or descriptions of P1's and P2's behaviours.

      Delete
    2. Patric,
      Thank you for the extensive reply!

      I guess my first thought is that the "objective perspective", as you put it, can contain the "subjective perspective". However, the opposite cannot occur. That is, with enough observation I can both see what is "really" happening and what the participants "think" is happening. However, if you start with the "subjective", you never quite escape into the "real", and that is a problem.

      "The assumption in the quotation is that P1 is wrong in her perception of P2's behaviour, and this will gain P2 in future hands. If P1 is correct however (and the audience wrong), it will gain her in future hands. Pragmatically then, an objective perspective does not hold any consequences for behaviour, only the subjective perspectives of the situated agents."

      But I think all professional poker players would agree that, given the opportunity, they would rather know what cards the other player had, because it would help them win more money in the long run. That is, I think most people would rather keep their opponent in the dark, but gain the "objective" perspective for themselves. This was a major concern of pro-players when the hole-cams were first introduced. So the objective perspective seems desirable, even if it is not obtainable at particular moments.

      "In poker, the perspective of the audience is indeed intriguing, and I believe it to be so because we can see players over-evaluating their own hand, under-evaluating their own hand, over-evaluating others' hands, under-evaluating others' hands, misreading behavioural cues, or accurately perceiving them"

      Yes, exactly! We can see those things. All you need to be a psychological realist is to take that claim seriously.

      "My answer to this is most often that an objective perspective almost always reveals many more options to describe behaviour from than a subjective one -but it doesn't mean that it is necessarily more accurate."

      Well... we can have a very interesting discussion here about "accuracy" in the world of a psychological realist world, but maybe that should wait for another post. The fundamental metaphor that I like to work with in realist philosophy (phrased most bluntly by Holt, but bequeathed to me by my post-doctoral mentor, Nicholas Thompson) is the metaphor of a Point of View. The world indeed appears different from different points of view, but it would be weird to assert that one point of view was "more accurate" than another in a grand metaphysical sense.

      ".... In this way, the subjective perspective is taken into account and eventually provides for an objective perspective, rendering the distinction between them fuzzy.

      Yes, yes!

      Thank you for all the good material to think about!

      Delete
    3. Eric Charles,

      Thank you for the time and effort you put into reading it, I do try to be concise but I guess your post sparked many thoughts =), also, thank you for your reply.

      I think I have a better understanding now of what you mean by what is "really" happening and what participants "think" is happening. Am I wrong in assuming that P1 even can believe she is bluffing, but, because of reason X, Y or Z, neither P2 nor the audience would consider it a bluff? And this is what would be central to the hand/game? The one thing I can think of is that, isn't this true for all cases of bluffs in poker? I.e. it is considered a bluff until another player gets wise? Would it then depend on the emotion/body language management skills of P1 and observational ability/specificity of P2, situational/context-dependent factors, if either perspective is more revealing than another?

      "So the objective perspective seems desirable, even if it is not obtainable at particular moments."
      Yes, absolutely, the temporal aspect is central here, I cannot see other players' cards at the time that I would want to, and this can be seen as a temporal constraint necessary for poker to have any value to players at all.

      I can't help but be reminded by Gigerenzer here, a wider perspective in a poker game -assuming that this perspective contains the information of other players' cards- gives relevant information to the player. One could argue that a wider perspective need not always be advantageous, it would then depend on the specific information that becomes available in this wider perspective compared to the information that the players have for themselves. If a wider perspective contains information not relevant, then it is probably more likely to mislead than aid.

      Agreed, I should not have used the word accurate here, it takes us down a garden path. I should have written valuable, or informative, or directly informs, because the gist was meant to reflect what I write in the above paragraph -that we most often assume a wider objective gives more information and more information leads to more grounded and informed decisions -when in actual fact it relies on which extra information is gained, if it is not relevant information then we do not gain very much from a wider perspective/more information.

      I appreciate that, thank you, I look forward to future posts here on Fixing Psychology, it is one of my favourites, thank you for sharing your ideas.

      Delete
    4. Patric,
      I think, from any perspective, whether or not one person is bluffing is independent as to whether or not the opponent is wiser. Admittedly, the bluff probably stops having its desired effect when the opponent gets wiser, but lots of people continue bluffs long after that point (in poker, and in "real life").

      It would be interesting to try to apply Gigerenzer's "relevant" criterion here. "Relevant to what?" is the obvious next question. If I am trying to evaluate what to do in a typical hand, I certainly would want to try to get a feel for whether or not my opponent thinks she is bluffing. Under some circumstances, I might also want to estimate odds that she is mistaken about whether or not she is bluffing.

      Part of my fascination with the statistics of poker is that information not particularly relevant to this particular hand, can still be relevant to higher-level aspects of the long-run game. If I am trying to become a better player myself, it is crucial that I become more accurate at knowing where I stand relative to other players, i.e., whether I am in a position to bluff or value bet.

      Of course, one never becomes perfect at these skills, and there is a whole other discussion to be had about putting your opponent on a "range of hands" rather than a particular hand.

      Delete
    5. Eric Charles,

      I apologize for not returning a response sooner, I usually get an email. Just wanted to add; understood, and agreed ;). I look forward to reading your recently tweeted articles/papers, sounds fascinating!

      Kind regards

      Delete
  2. For example, this person [who has knowledge of the hole cards] knows that Player 1 is not bluffing.

    This seems like a strange way to put it. The person watching knows both players' cards, whereas Player 1 only knows that she has a pair of 7s. As far as Player 1 is concerned, Player 2 could have a pair of 8s, Js, Qs, some sort of three of a kind, a straight, etc...

    I think the strangeness lies in defining bluffing with respect to complete objective knowledge rather than the partial knowledge of the player doing the bluffing (or the not-bluffing). If Player 1 thinks she has a worse hand than Player 2, it seems more reasonable to call the bet a bluff than to not do so. And if Player 1 thinks she is bluffing, then I imagine her behavior and physiology would be indistinguishable from cases in which the complete objective knowledge and her partial knowledge match (i.e., she's truly bluffing, according to the definition given here).

    ReplyDelete
    Replies
    1. Noah,
      It seems you have honed in on the exact point, but are unsure which position you prefer. As I stated, there is a game of limited information going on, so that description has certain merits. It is just limited.

      Two quick questions:

      1) Let us assume you are a beginner trying to improve your poker skills. After a long conversation, analyzing your hands in a series of games, I might say, "It seems like you think you are bluffing in a lot of situations where you are not bluffing." You reply, "That is impossible, bluffing is a state of mind, and so I cannot possibly be wrong about whether or not I am bluffing." What on earth am I supposed to say next?!?

      2) There are many problems with polygraphs used as "lie detectors". One problem is that if people do not know they are lying, then there will not be the accompanying change in vital signs that the readers hope for. However, I think we all agree that the person is lying depending on whether or not their statement is accurate, not based on whether or not their physiology changes. We might bicker over whether the person was trying to lie, or whether they intended to lie, but I think we agree that it is a lie if they make factually inaccurate statements. Shouldn't it be the same in this case?

      Well... that is really the point isn't it? There are advantages to looking at things this way. For example, if you are a player who does not know when you are bluffing and when you are not bluffing, then you will have a lot of trouble improving your poker game.

      Delete
    2. Thanks for the response. I'm not sure if it's an issue of preference for me, since both positions are (or seem, anyway) valid, just from different perspectives. That is, it's probably worth distinguishing bluffing as a state of mind and bluffing as a matter of fact.

      It seems like the non-psychological part of poker plays an important role here. If Player 1 knows all the relevant probabilities, then she can make a statistically optimal decision to bluff or not - she can know if she is more or less likely than Player 2 to have the best hand, and she can, in the latter case, decide to bluff, or, in the former case, make a value bet. But it's not clear to me how she could ever know, for certain, whether she actually has the best hand.

      So, it's easy to imagine cases in which Player 1 state-of-mind bluffs even though she isn't matter-of-fact bluffing, or where she both state-of-mind bluffs and matter-of-fact bluffs, or either of the other combination of values for the two types of bluffing.

      Of course, in the general case, the psychological part of poker is also important, so whatever Player 1 knows about Player 2 also enters into the equation (e.g., Player 1's knowledge of Player 2's propensity to bluff, 1's knowledge of 2's knowledge of 1's propensities, etc...).

      But even complete knowledge of other players' psychologies won't eliminate the probabilistic, non-psychological component of the game.

      By the way, I've been reading your blog for a few months now, and I enjoy it quite a bit. My background is in linguistics, math psych, and cognitive science (i.e., not ecological psychology, for the most part), and I've been finding it very interesting to read Fixing Psychology (as well as Andrew Wilson and Sabrina Golonka's blog).

      Delete
    3. "But it's not clear to me how she could ever know, for certain, whether she actually has the best hand."

      Well... that's what makes it a game, right. It would be trivially easy for her to know if she had the best hand. All she has to do is see her opponent's cards. The rules, however, discourage that.

      To rephrase your next paragraph: "So, it is easy to imagine a situation in which Player 1 thinks she is bluffing, though she isn't actually bluffing, or where she is doing both, or neither."

      You are right about that, of course, and reading "tells" (including betting sizes), becomes interesting in part because of the many potential psychological things that can be happening. There is a way that Player 1 can act when she is "thinking she is bluffing", as well as a way she will act that is "wanting you to think she is bluffing", etc. The psychological side isn't eliminated in the realist perspective, it exists as potentially observable things that people are doing.

      I'm glad you enjoy the blog! Thank you for the comments.

      Delete
  3. A great and much beneficial blog in regards to dealing with Psychological problems and issues impacting mental health. Psychiatrist in New York City

    ReplyDelete
  4. Are you familiar with Howard Rachlin's work? He embraces the use of mental terms, but uses them to describe overt patterns of behavior over time. Your line about being able to see the minds of others reminded me of some of his writing.

    Good post.

    ReplyDelete
  5.       While I was searching on the net I came across your website and I found most interesting to read. The language used in the articles is easily understood. I will continue to make use of this site in my future also.
    casino

    ReplyDelete