This user hasn't shared any profile information
Posts by Dan
A few weeks ago I was trying to teach my niece and nephew how to play ping pong. For anyone unfamiliar with the game, you use little rubber-covered wooden paddles, a bouncy ball, and a table with a net in the middle, and you get a point if you successfully hit the ball over the net with the appropriate number of bounces (once on each side for the first hit in the round, only one time on your opponent’s side thereafter) and your opponent is unable to do the same back to you. As is natural while learning a new game they were very interested in figuring out how well they were doing, which they attempted to ascertain by asking “did I get a point for that?” a lot. I told them not to worry so much about whether they were getting points, but to just focus on their technique and getting the ball over the net with the appropriate number of bounces. I did this because I couldn’t remember at the time whether there was some interaction between winning points and who served in a round, but more importantly because the “and your opponent can’t return it” part is such a huge factor in whether you score that it takes the focus of what you are doing, which is only realistic to focus on once you have solid enough fundamentals that you have a decent amount of control. If they wanted to be truly competitive they’d eventually need to learn how to take their opponent’s capabilities into account (e.g. I have decent accuracy with my backhand but my forehand is terrible, so trying to get the ball to a place where I have to use my forehand to hit the ball is a good way to get a point against me), but for casual play and practicing basic skills they should focus on what they were doing and how that was interacting with the ball and the table. This casual interaction was enjoyable for all of us, especially since I’m not that great a player myself, but if I was more skilled and was looking at the game very competitively their capabilities would have loomed large: either I’d be getting hollow victories by beating opponents who couldn’t play or I’d be patronizingly letting them win. It occurred to me that how much attention we’re paying to our assessments of other participants has a big impact on the feel of a game, and I think it relates to an important concept for RPG design.
The term I’m picking to talk about this is Posture. Posture is about where and how participants stand with respect to a game (where I’m using “stand” metaphorically). The thing that most strongly controls this posture is what we’re paying attention to or what we care about on a moment-to-moment basis. If I have a competitive posture in ping pong I care about winning and I need to pay attention to the capabilities of my opponent in addition to the bouncy-ball-physics element of the game. RPG mechanics can lead us to focus on things in ways that affect our Posture. For example, both Fate and Mouse Guard have a mechanic where the GM sets a “target number” that players must “roll against” to see if they succeed or fail at a task, but there are big differences between them. In Fate, the GMing instructions tell you to set the target numbers relative to a character’s skill so that you can orchestrate the experience of play: if you want the player to have a chance to “showcase their character’s awesomeness” you give them a low difficulty number, if you think that have “too many” points of a spendable resource you give them a high difficulty number so they’re likely to spend them to succeed. By contrast, in Mouse Guard the GM compares the fictional situation as-described against a series of factors in the rules for each skill: finding water in the wilderness is always the same difficulty, no matter whether the game been tense or leisurely, or whether the player is excited or reluctant. Fate wants the GM to focus on their moment-by-moment assessment of the player: what target number should I give them to provide the psychological jolt I think they need to get the best experience? Mouse Guard wants the GM to focus on filtering the fictional situation through the game procedures, presuming that faithfully engaging the process will produce the desired fun. Since the games put you in different postures for making this superficially-similar decision there is a difference in the subjective experience of play. And, since the people are a vital component of an RPG-in-operation, giving them different subjective experiences also impacts their behavior, which makes the games different in play.
Getting participants into the proper posture can have a big impact on RPG play. The GMing rules in Apocalypse World, for example, are largely about getting the GM into a posture so that they can comfortably make the decisions they need to make in that game. (There’s also a rule about player posture: “play your character like they’re a real person”, but it’s less conspicuous and frequently ignored). A game like Primetime Adventures that’s built around character-conflict tends to need players to lean in aggressively and can flounder if players adopt a more passive don’t-rock-the-boat posture. In a sandbox-y “dynamic world” game the GM might be flummoxed if the players adopt an accommodating “we’ll do what we think you want us to do, because we want to experience the story you planned” posture. Players bringing goals and ideas from other games can often give them an incompatible posture for a game: if a game expects you to be a self-directed goal-seeker but you’re playing with a performative improv-actor “I’m trying to entertain everyone else” posture the game may break down.
While I’m saying that I think Posture is an important concept, I don’t want to give the impression that I think it’s simple or one-dimensional. In general, the posture players take has a huge impact on which directions they can experience tension or pressure without suffering excessive strain, which means that properly orienting the players’ posture is an important part of game design. It seems to me that in the RPG design space we’re just beginning to learn how to do this effectively, so I believe it’s a fruitful area for further thought and exploration.
I recently read a psychology paper and was deeply struck by the implications it might have for a wide variety of tabletop roleplaying game and RPG design issues. If I’m right, it speaks to issues of immersion, a difference between RPGs and so-called “storytelling games”, the social power dynamics between players and GMs, the reason that many of the techniques used in Apocalypse World work well for many people, and more. The paper is Construal-Level Theory of Psychological Distance by Trope and Liberman, and it summarizes and consolidates findings from many individual experiments into a unified theoretical framework. I’m going to give a brief overview of the theory, and then talk about some of the lower-level details and experimental findings, and then talk about some of the links to RPGs that I find so exciting.
Summarizing the Theory
The basic idea of Construal-Level Theory (CLT) is that we humans are generally capable of interpreting the same thing through different “construals”, either at a low, concrete level which includes lots of contextual and peripheral details, or at a high, abstract level, which will be schematic, decontextualized, and focused on central, prototypical features. For example, we might think of a baseball bat at a concrete level as “a Louisville Slugger”, while we might think of it as “sports equipment” at a high level. Naturally, the same thing can be abstracted in different ways based on which features are considered central for the way the brain is trying to use it, so another abstract construal of a baseball bat might be “a bludgeoning weapon”. Generally speaking you can nearly always get even more abstract (weapon, hand-held object, thing) or more concrete (an autographed Louisville Slugger, the bat used to hit the game-winning run at the World Series last year). CLT suggests that operating at a high or low level of construal is a mindset, vaguely analogous to shifting gears in a car: the way we process one thing tends to be sticky and influence the way we process the next thing.
It’s important to note that CLT doesn’t frame this in terms of laziness or quality of representation, since high- and low-level construals are better suited to different kinds of cognition, neither one is “better” than the other. For example, high-level construals lose the peripheral details but they add meaning and the ability to put things into systems: no amount of zooming-in on the bat’s woodgrain will give you insight into how it’s used in a baseball game. It’s beneficial to be able to see both forests and trees, but often hard to do both at once.
CLT further proposes that human brains have an association where we tend to use high-level construals for things that are far away and low-level construals for things that are close. If you think about this in terms of visual perception and physical distance you can see why this association makes intuitive sense: When something is far away we may not be able to make out all of the details, and the contextual ones may change if and when we actually interact with it, so in general it makes sense to use high-level construals for distant objects. And when something is close we have access to all of the details and may not have strong reasons for abstracting it in a particular way so we tend to use a low-level construal. But since this association is somewhat automatic we do it even when those justifying factors aren’t true: even if we could access all of the contextual details of a distant target we still tend to use a high-level construal, and even if it might be beneficial to think of something close in an abstract way we’ll tend toward the more concrete construal. So, for example, we’re likely to think of a distant car in an abstract way even if we’re familiar with the minute details of that particular make and model.
CLT notes that this phenomenon seems to be consistent across multiple dimensions of psychological distance. So, for example, we tend to abstract to the same degree when thinking about something that happened long ago as we do when we think about things that are far away. In addition to temporal and spatial distance, CLT also says that hypotheticality (e.g. a near certainty vs. a remote possibility) and social distance (a close friend vs. a distant stranger) work the same way.
The final piece of the CLT puzzle is that the association is bidirectional: we use psychological distance as a guide for what construals to use, and we use the construal mindset we’re using as a guide to what sort of distance something has. Something construed more abstractly will tend to feel more distant, something construed more concretely will tend to feel closer. This struck me as highly relevant to roleplaying games, where we often seek to feel close identification with characters and want a sense of here-and-now experiencing of fictional events unfolding rather than merely knowing that they happened. The research tends to be much coarser-grained than this, such as comparing next week to next year or this city to a city on the other side of the country, so extending the ideas to the fine-grained world of RPG experience requires some speculation, but I think that speculation is warranted. As I’ll explain below, I think there are several successful RPG design features and common RPG breakdown modes that map to predictions from CLT, which make me think this could be a very useful tool for understanding RPGs.
Differences Between High- and Low-Level Construals
The experiments have noted several differences between operating with high- and low-level construals. Since the relationship between construal level and distance is bidirectional, sometimes the high- or low-level construal condition is used as the input to the experiment and sometimes it’s used as the measured output. Here are some of the things that the research associates with the different levels of construal:
Seeing forests or trees: In several experiments, working with big-picture “gestalts” tends to be associated with distance, and nearness lends itself to working with constituent details. For example, in images where there’s an overall gestalt effect made up of smaller individual elements (such as title image for this blog post, where there’s a big RPG made up of little CLTs), people in the distant condition tend to more quickly and reliably detect the large gestalt image, while people in the near condition tend to more quickly and reliably notice constituent details.
Broad or narrow categories: High-level construals tend to be associated with fewer, broader categories while low-level construals tend to be associated with more, narrower categories. In experiments where subjects were asked to categorize a list of items into as many groups as they felt appropriate, subjects in distant (high-level) conditions tended to create fewer groups with more members in each while subjects in near (low-level) conditions tended to create more groups with fewer elements in each group. For example, when asked to break a video sequence into “meaningful actions”, subjects that believed the video was shot nearby tended to break it up into more, shorter snippets than those in the distant condition.
How vs. Why: When examining actions, focusing on “why” or the consequences of the actions tends to be associated with high-level construal while a focus on “how” or the feasibility of the action tends to be associated with low-level construal.
Situation vs. Traits: Low-level construals tend to include consideration of situational factors while high-level construals tend to be associated with things that look like stable properties, e.g. your friend didn’t respond to your e-mail because she has many demands on her time, a stranger didn’t respond to your e-mail because she is inconsiderate.
Weighting of attributes: When operating at a high level of construal people tend to put much more weight on some attributes of choice than others, while operating at a lower level tends to lead to attributes being more evenly weighted.
Alignable vs. Nonalignable attributes: When people are comparing things in low-level conditions, traits that are easily compared between items tend to heavily influence preferences (e.g. this meal has less calories than that one) while high-level construals more easily allow preference formation between nonalignable attributes (e.g. this meal is from Italian cuisine, that one is from Vietnamese cuisine).
Generating exemplars of a category vs. naming a category for an example: In a very direct manipulation, having subjects generate subordinate members of a category (e.g. “A type of soda is ________”) tends to put them in low-level construal mindset and finding superordinate categories for objects (e.g. “Soda is a type of ____________”) tends to put them in a high-level construal mindset.
Values: Things like values and morals tend to be abstract ideas, and seem to show the same associations with distance, e.g. people tend to make harsher moral evaluations in distant conditions than in near conditions.
Types of Near and Far
Spatial Distance: Different experiments have manipulated distance in physical space in multiple ways, both by linking events to real geographical distance by telling people where events are taking place, or even by getting them to plot two points on a Cartesian plane that were close together or far away from each other.
Temporal Distance: Remembering the past and predicting the future seem to be roughly equivalent in terms of greater distance being associated with greater abstraction, i.e. next week and last week are about the same, and they’re closer than the distant past or far future.
Hypotheticality: Several experiments have shown similar construal-level effects to other distance dimensions by interpreting low probabilities as far (a remote chance) and high probabilities as close (a near certainty).
Social Distance: The amount of “social distance” seems to have similar effects, e.g. self vs. others, people in similar organizations compared to different ones, etc. Social power seems to have an effect as well, since priming people to think about high levels of social power tends to be associated with abstract construal and priming people to think about having low social power tends to be associated with concrete construal.
Putting it Together with Some RPG Examples
In Dogs in the Vineyard, when players are engrossed in a blow-by-blow conflict they’ll tend to escalate violence or use weapons because the rules make those options seem like “easy” ways to get extra dice. After the conflict ends and players are encouraged to look back at what happened in order to add new traits to their characters they often become uncomfortable when they realize how their in-character actions diverge from their contemporary morals. The short action snippets encouraged by the mechanics and the need to produce examples of “how” you are using your traits tends to lead toward a low-level mindset, so the feasibility of using violent escalation matters more than the desirability of using violence (and in the heat of the moment the “why” of what’s at stake in the conflict overall becomes less and less relevant). By contrast, in the post-conflict Fallout procedure, looking backwards at the gestalt entirety of the conflict makes it easier to process the events thematically, morally, or in terms of overall character, often leading people to morally disapprove of actions they themselves decided the character should take.
Sometimes players choose to play Dogs in the Vineyard in a way that focuses on stable traits of personality, such as violent tendencies, religious zealotry, etc., instead of engaging in-the-moment with the situations before them and the psychological pushes and pulls cooked into the design of the game. These two approaches seem to hew closely to different levels of construal, and people tend to report different subjective experiences of play from the two approaches (the people in the second group are sometimes described as playing caricatures rather than characters in order to keep their emotional distance).
Apocalypse World seems to have multiple techniques the lead people to operate largely with concrete construals. Rather than tying actions solely to a few global attributes, AW gives each play an abundance of “moves” through which they can affect the world (more, narrower categories). Focusing on “how” a character is making a particular move, or what particular thing they’re doing that’s in the category of the move, also tends to push toward a low-level construal. The GM is encouraged to dynamically scale the pace of resolution, breaking things up into a zoomed-in action-by-action level when they want especially strong connection with the fiction or characters. The relatively high probability of succeeding with any given move may also lead to low-level construals. Like DITV, AW also has several mechanics that sometimes ask players to step outside of their concrete interactions and interpret things in a more abstract way, such as with the Read a Person questions like “what does your character wish I’d do?”. While this is good for gaining insight into characters, it can sometimes be difficult for players to get back into a flowing interaction after answering them, perhaps due to different mindsets being conducive to the different types of processing.
Choosing Continuous Coverage Instead of Scene Framing: Some games give players the choice to set up their own scenes, but many people use this power to simply continue what was happening in the previous scene, sometimes seemingly getting bogged down in “logistical details” like how characters get from one place to another rather than jumping forward to get to “important” scenes. A focus on “how” is what CLT predicts for people operating in a concrete mindset, while figuring out what the most thematically appropriate or interesting-for-the-story next scene should be is something that involves relationships between abstract ideas and is therefore the kind of thing that’s easier to do in a high-level mindset. If people are having a very here-and-now experience with a scene and feeling very low distance between themselves and their characters then they’re probably going to see things through the concrete mindset and probably have a harder time shifting gears to do good “scene framing”.
“Storyboarding”: An experience that people sometimes report in games with explicit-stake-setting conflict-resolution mechanics is that they’ll elaborately negotiate out the possible consequences of a conflict, roll to determine which outcome happens, and then move forward with the game knowing which outcome occurred without ever having the here-and-now feeling of experiencing the scene. CLT suggests that focusing on the “why” of conflict stakes will shift people toward abstract construals and therefore increase their associated sense of temporal distance, i.e. move them away from “now”.
GM/Player Social Dynamics: Some games or playstyles explicitly vest the GM with special social authority (e.g. the ability/responsibility to determine who is “playing well”), sometimes this happens as a byproduct even in games that desire egalitarianism despite the asymmetry in roles. If being “a player” in a game involves mostly being in the concrete-construal mindset and being “a GM” involves more abstract interpretation and categorization then the perception that a GM has more social power than a player may reinforce that (or be a byproduct of it).
Some Questions That Arise
Since several of the experiments show distance-related effects merely from telling the subjects that they’re dealing with near or distant places I think it raises some interesting questions for RPG settings, since we often set our games in times or places that aren’t here-and-now. Do “modern” games tend to feel more immersive? Do games set in exotic locales make it harder to feel a strong sense of here-and-now? Although I don’t have much personal experience, it may not be a coincidence that players of urban fantasy games like White Wolf’s Vampire seem to prioritize immersion and strong character connection and also tend to pretend that the fictional events of the game are secretly happening in their own city right now. What about the fantastical element? Most RPGs have some “genre” element of fantasy or science fiction, and most “medieval fantasy” games are set in other universes rather than historical earth. Is that conducive to low-level construals, or a distraction? In the paper, Trope and Liberman say:
We argued that because high-level construals are broad, they bring to mind more distant instantiations of objects, and because low-level construals are narrow, they bring to mind more proximal instantiations of objects. It is also possible for construal level to affect the psychological distance of objects through metacognitive inferences (N. Schwartz & Clore, 1996). People may interpret their low-level construal of an object as indicating that the object is close and their high-level construal of an object as indicating that the object is distant. This metacognitive inference of distance from construal level might involve a more complex attributional calculus when one or more other distances are known. Specifically, the construal-based inference that an object is distant on any given dimension will be discounted when the object is known to be distant on another dimension. Correspondingly, the construal-based inference that an object is distant on any given dimension will be augmented when the object is known to be proximal on another dimension. For example, one would attribute a detailed construal of a meeting with a friend to a relatively close relationship with that friend when the meeting is known to take place in the distant future rather than the near future. Thus, direct implicit associations among different distance dimensions generally result in positive relationships among those dimensions. However, when inferring distance from construal, adjusting the inference of distance on one dimension for distance on other dimensions may result in a negative relationship among those distances.
Could fantastical elements create a plausible place for players to shunt unwanted perceptions of psychological distance, thus facilitating feelings of closer distance on other dimensions?
Also, it seems like the experiments that used Cartesian coordinate-plotting to prime distance might have implications for games that make extensive use of physical maps, from the maps-and-minis tactical combat games like D&D 3.5 and Pathfinder, to map-focused games like The Quiet Year or Chronicles of Skin, to games that use relationship maps to represent character relationships. The physical distance priming involved in working with these maps could be having an interesting and perhaps previously-unrealized impact on play.
Some Areas For Further Exploration
Several of the distinctions and manipulations associated with the different construal levels seem like useful guides for some common design decisions. For example, the differences between few/broad categories and many/narrow categories in these experiments might help a designer choose between a large skill list or a short pool of attributes to achieve different effects. The research might also suggest different effects for different types of player-authored traits, e.g. broad dispositional traits like “clever” or “rigid” would be expected to guide players toward abstract construals while traits that are more likely to interface with the situation like “well-armed” or “quick” might guide players toward more concrete construals.
There are also intriguing structural possibilities to explore. For example, Apocalypse World, like many other RPGs, uses character attributes to modify the probability that moves will succeed or fail, with the results of those actions generally fixed on a per-move. However, it could be interesting to create an AW hack in which we switch things around so that different moves (i.e. different types of character action) have identical high likelihoods of success but different levels of impact based on character stats. This change might make a player’s choice of move more revelatory-of-character by making that decision more intuitive and in-the-moment if the effect of different stats is “hidden” in the consequences (which people are more prone to discount when they’re operating with a concrete mindset) rather than putting a thumb on the scales of feasibility.
Hopefully I’ve done a good job of summarizing the research, but obviously that involves employing some editorial judgment in my choice of what to say and how to say it, so wherever possible I’ve provided links to the source material (with citations in the alt-tags). As I said earlier, my linkage of some of these RPG design issues to this theoretical framework involves some speculation, so we shouldn’t necessarily assume that the inferences I draw are rock solid. Furthermore, much of the social psychology research that feeds into this theory involves somewhat coarse-grained between-subjects statistical effects, and the effects measured in the experiments are not always strong night-and-day differences, and not all of the experiments have been widely and reliably replicated. And, like any scientific theory, it’s a model that attempts to account for the observed data, not necessarily anything that can be directly confirmed. Additionally, the experiments rarely speak at all to issues of individual variation (e.g. even assuming the theory is roughly correct, are some people able to switch back and forth more easily than others?). And, since most of the experiments are about merely establishing a difference between near and far, abstract and concrete, the research doesn’t give us anything like nicely parametrized curves which we can plug directly into some kind of game design formula.
Those caveats aside, though, I think it is intriguing research that suggests relationships and general patterns of human thought and behavior that we may be able to leverage to make better games.
I recently learned about the “Benign Violation” theory of humor which some researchers are using as a framework for studying the psychology of humor. The basic idea is that we find something funny when it’s both 1) a violation and 2) benign. So tickling is funny because it’s an attack (a violation) but harmless (benign). Puns are funny because they violate expected usage but in a way that carries no negative consequences. Self-deprecation is funny because it is saying something negative (a violation) but in a way that we can tell isn’t sincere (benign).
It’s an interesting theory because it predicts both what is and what isn’t funny. Tickling by a stranger, for example, isn’t funny because we don’t know if it will be harmless. And self-deprecation isn’t funny if we suspect the person saying it is serious. It suggests that to be funny something needs competing interpretations, one where it’s OK and one where it’s not OK, and they need to be relatively matched in intensity: something that’s a trivial violation in an entirely benign situation isn’t funny, nor is a major violation with a figleaf, at least according to this theory. Intuitively it sounds like a pretty compelling argument to me, and it may be a useful framework for exploring the design of comedic RPGs, a famously fraught endeavor where plenty of games are maligned as “funny to read, not funny to play”, or for keeping unwanted comedy out of dramatic or gritty games.
It seems to me that many classic funny RPG situations map to this theory: The dragon opens its jaws and lunges forward to [rolls to hit] bite you for [rolls damage]… 1 point of damage [laughter ensues]. In the hit/miss framework it’s a violation, but in the damage framework it is benign, so we are amused. One of the most reliable design techniques for comedy RPGs is to have players describe their own character’s failure: since they failed it’s a violation, but since they have control over the situation it’s benign. If a player makes an out-of-character joke it can be funny, and a pretty reliable technique for eliminating that humor is to make it consequential in the fiction by acting as if the character rather than the player said it (i.e. it’s no longer benign). The gold-standard of funny RPGs, InSpectres, seems to fit the pattern, too: it instructs you to play “normal” characters rather than outlandish ones, which would seem to help with interpreting the supernatural events of the scenario as “violations” but the relatively low stakes of the situation make them benign (these supernatural situations are so commonplace in the in-game world that resolving them is a viable franchise business model, and a session is about whether players satisfactorily complete the job to bring in money to keep their business afloat).
While this framework seems appealing at first glance, a deeper analysis is probably warranted. It would be beneficial to go beyond cherry-picking examples that seem consistent with the theory and look at a spectrum of games and see how well the funny/not-funny predictions hold up (analyzing Cthulhu games could be a particularly fruitful way to approach this, given the diversity of approaches in that genre and the fact that they’re sometimes played seriously and sometimes for laughs). Additionally, to be useful as a design tool we’d need to see if we can translate the concepts into game design elements: What counts as a violation? What makes things benign? How do we evaluate the relative intensities? How does wanting games to be “fun” interact with the “funny” that we’d include or exclude when employing this framework? Attempting to use this theory as a prospective guide to designing a new funny (or reliably unfunny) game could also be a fruitful avenue of exploration.
One of the memes floating around in the tabletop RPG design-o-sphere is that it’s desirable to reduce the “social footprint” of games. The thinking goes that busy people have a hard time fitting gaming into their lives, so when games require things like learning rules, regular attendance at scheduled sessions, outside-the-session prep-work, etc., it makes it less likely for the game to happen. While this argument is compelling, we shouldn’t assume that reducing barriers to play is a purely beneficial strategy that has no tradeoffs. In addition to affecting which tools are available from the game designer’s toolbox, the “inconvenience” of getting a game to happen can itself have an impact on play. I recently read a psychology research paper that illustrates an interesting phenomenon.
In the experiment I want to highlight, the subjects were asked to participate in a short test to measure performance on some mental tasks. Different subjects were scheduled to take the test at three different times (selected to be mildly, moderately, or extremely inconvenient). After the students were told about the test and when they’d be taking it (without knowing the scheduled time was a variable) they filled out a survey indicating how important or interesting taking the test would be, and how satisfied they expected to be once they completed it:
As you can see in the chart, subjects in the late-but-not-too-late condition rated taking the test as slightly more important and meaningful to them. The paper puts forward the theory that when we are exposed to short-term costs between us and our goals, our brains use techniques like magnifying the significance of the long-term goal to make sure we get past the bump in the road, but we only do this within the realm of the possible and don’t bother if the short-term cost seems too high. Since the subjects subconsciously anticipated difficulty staying motivated to perform at the moderately inconvenient time, their brains helped them out by deciding the overall task was comparatively more important than their less-inconvenient-time or it’s-so-late-it’s-a-lost-cause peers who were evaluating the exact same task. And here’s the even more interesting result:
You’d probably expect average performance on the test to decline based on the later times (presumably people are more tired later), but the average performance by the moderately-inconvenient-time group was actually slightly higher than for the earlier group. Using some statistical analysis, the experimenters say that there was a negative effect on performance relative to time as you’d expect, but there was also a positive effect on performance relative to how important and significant the subject considered the test (presumably they try harder). Since the moderately-inconvenient subjects thought the task was more important their increased motivation compensated, or more than compensated, for the lateness of the task. Obviously low-inconvenience is better than high-inconvenience, but comparing low and moderate inconvenience may not be so straightforward.
Naturally all the standard caveats apply about the risks of generalizing from an experiment like this and applying it to a different field like RPG design, but I think it’s worthwhile to consider whether trying to minimize the “social footprint” might risk throwing the baby out with the bathwater. When there’s a little resistance to making a game happen the players are probably playing the game with a slightly different mindset compared to players who face no inconvenience, and that can easily have an impact on what techniques, systems, procedures, etc., will work well for those players. This isn’t to say that designers shouldn’t consider the importance of right-sizing the “social footprint” of their games, just to caution that a simplistic “less is better” strategy may not be optimal.
Most aspiring tabletop game designers face a challenge in finding external playtesters for their games. While external playtesting is essential for confirming that that the game works as desired when the designer isn’t there to facilitate play, groups willing to playtest an unknown designer’s game are few and far between. Naturally this leads to pondering whether there are incentives that could convince people to playtest. In our society the “go to” incentive is usually money, but most aspiring designers intuitively grasp that they can’t realistically pay a playtester what they’re worth without taking their project deeply into the red.
Putting a decision in the “money” domain often changes the way people look at it. Consider this hypothetical: My car gets a flat tire and I say “Hey, will you help me change my tire?”. Most people, as long as they’re capable of doing so, would instinctively say yes. Now consider what would happen if I said “Hey, will you help me change my tire for fifty cents?”. Most people would instinctively say no, because their time and energy is worth far more than fifty cents. Offering a small financial inducement usually doesn’t stack on top of people’s natural altruism, it shifts the domain of consideration from a social context to a monetary one.
Of course, money isn’t the only possible inducement. Sometimes people reciprocate in the social domain by exchanging gifts such as bottles of wine, restaurant meals, or (perhaps more realistically for aspiring game designers) PDFs of games or game supplements. Although it’s common to offer the retail-version PDF of the final game to playtesters I’ve always had a philosophical hangup with that: since playtesting may reveal that the game has fundamental flaws that prevent it from ever being productized, and since playtesting may reveal that a particular group doesn’t enjoy playing the game, you can’t guarantee that you’ll have a final retail-version of the PDF that will be valuable to the playtesters, so does it make sense to use that as an incentive? I’ve long hypothesized that it might be easier for a designer to attract playtesters to a second game, since you could use free versions of the first as an inducement. But I just read a behavioral economics research paper that casts some doubt on that idea.
In the experiments they asked subjects to do a task in exchange for various incentives and measured how much effort they put in. In addition to testing with small and medium monetary rewards they also measured how people responded when they were offered small or medium gifts of candy, and additionally measured how they responded when the gifts were described with a clear monetary value, e.g. a “fifty-cent chocolate bar” or a “five-dollar box of chocolates”:
As you can see in the charts, low payment produced lower effort than no payment (the control condition) or nonmonetary gifts. But the same gifts had nearly the same effect as the equivalent amount of cash when their monetary value was clearly understood by the person doing the task. This leads me to suspect that giving people a free PDF which has an easily-ascertained retail price could be just as demotivating as offering a small amount of cash in exchange for playtesting.
It does suggest that offering other things that don’t have an obvious price, such as tchotchkes or perhaps game content that isn’t available in any other way, can be effective. An idea I’ve pondered is generating supplemental content for a popular game, such as a limited-edition playbook for Apocalypse World or a character class for Dungeon World, that could be used as an inducement to playtest a new standalone game. I haven’t pursued that myself since I don’t feel I have enough play experience with those games to create quality content for them, and the new content itself would need playtesting. Also, I suspect that if I tried to engage with those games with a mercenary “means to an end” mindset I wouldn’t enjoy playing them as much, and enjoying games isn’t something I’ve been willing to risk.
If I did have a valuable inducement to offer to playtesters I think it might also help bridge another of my hangups, which is that the “beggars can’t be choosers” effect makes me reluctant to push for playtesters to playtest well since I have trouble getting people to playtest at all. While I’m not usually a fan of purely transactional interactions, it does make it less socially demanding to set expectations when both parties are clearly and concretely benefiting from the interaction.
I recently read a short psychology paper that illustrates a point that tabletop game designers may want to consider. The experiment had two premises: First, that when we have a greater number of choices we’re more likely to find one that’s a good match to our preferences, so having more choices should make an individual more satisfied. Second, that when we have a lot of choices it has also several negative effects, such as requiring a lot of cognitive work to do the comparisons, highlighting the downsides of comparative choices by giving a basis for comparison, etc., so having more choices can make an individual less satisfied. The experimenters ran a trial asking people to evaluate a set of pens to identify the best one and then offered to let the person buy the pen they liked best. Here is a graph of their results:
As hypothesized, the two forces seemed to combine to create a “sweet spot” at which more people were satisfied enough to actually pay for a pen when they found one they liked in a some-choices-but-not-too-many situation.
Obviously not everything is a pen, different people may respond more or less strongly to competing psychological forces, and we should always be cautious about generalizing too much from a psychology experiment, but it doesn’t seem unreasonable to me that the inverted-U-shaped curve in this experiment may be a common phenomenon. If so, it points to a danger of an RPG supplement model that causes a game to include a monotonically increasing set of mechanical options like character classes, feats, powers, etc., because that might move the game out of a sweet spot as the number of choices accumulate. For example, issuing a series of “limited edition playbooks” for an Apocalypse World-style game might increase the odds that an individual player will find the perfect one for them, but it might also result in players agonizing over their choice of playbooks when play begins and make them less satisfied during play as they lament what might have been if they had gone with one of their rejected choices. I’m not trying to criticize the game design or marketing choices of any particular game here, merely pointing out that there may be easily overlooked downsides to embracing only a “more choices is better” strategy.
There’s currently no established consensus about how to develop a tabletop game once you’ve gotten past the initial “burst of inspiration” design stage. Most people more or less agree on the end goal, which is that they want a promising initial design revised until it’s “good enough” that it would be able to comfortably sit on a real or virtual shelf next to other published games (the issue of commercial vs. free publishing sometimes complicates that further). Most people intuitively expect that “playtesting” needs to be part of that process, and many people use “playtesting” as the blanket term to describe what they’re doing. Different models of how this development process should proceed will lead to different strategies and tactics for success. In this blog post I want to describe two different ways to look at it.
First, I’ll describe what I’ll call the “funnel” model. This model starts from the premise that a designer needs feedback about the gameplay in order to get to the game to a publishable state. And then the thought process goes: In order for the designer to get feedback, someone needs to hear about the game, get the playtest materials, read the materials, decide to play, organize a playtest, run that playtest, generate feedback from the playtest, send that feedback to the designer, and the designer uses the useful feedback to improve the game. Then you apply some insight and notice that there transitions between the steps: Not everyone in the world hears about the game. Not everyone who knows about it will get the materials. Not everyone who gets the materials will read them. Etc., etc. You can diagram it like this and notice that the “dropoff” at each level creates a funnel shape:
Since you’re trying to increase the bottom level (useful feedback) until it hits something analogous to a critical mass, the diagram suggests several strategies: Maybe you can widen the top of the funnel by talking about your game in lots of places. Maybe you can address the dropoff between “get the playtest materials” and “read the materials” by giving the playtest document a snazzy layout that makes it very readable. Maybe you can develop notoriety in another field, like blogging or podcasting, thereby expanding your audience. Maybe you can develop a great “elevator pitch” so that hearing about the game is very likely to lead people further down the path. Maybe you can develop a reputation for being smart or trustworthy in another medium (or with previous games) so people are likely to give you some benefit of the doubt to cover transitions where they might otherwise drop off.
Maybe you’ll also notice that the variables aren’t as independent as the diagram may make them seem, and you’ll figure out a way to simultaneously optimize some of the dropoffs. For example, perhaps you’ll conclude that one reason that people might hit the “read it but didn’t want to play” barrier is because they don’t want to play games in genres they dislike. You can avoid that problem by primarily communicating with fans of the game’s genre instead of a general audience: the top of the funnel is narrower, but you won’t have to deal with genre-rejection at the lower levels.
This funnel model has a lot going for it. It tells a story about how to succeed that sounds consistent, and it suggests tactics and strategies for how to make progress towards achieving that success, which is what you need from a development model. It also looks very similar to the way some people envision marketing a completed product, so solving the problems that this model tells you to solve may be helpful for also achieving success with the published game. It also has some downsides which I’ll get to later.
A different model starts from the premise that a well-designed game product will generate enjoyable play, people who experience enjoyable play will want to tell their friends about it, some of those friends will want to also have that enjoyable experience and seek out the game product, they’ll enjoy playing it, and thus you have a self-reinforcing word-of-mouth marketing strategy. Step one in that strategy is having a well-designed game. Few people are lucky enough that all of their ideas are perfect in their initial form, so in order to get a well-designed game you need to develop it from your initial conception into something that works. In order to do that you need playtests to generate data about what is or isn’t working, and you use that data in a successive-refinement game design/development and playtesting cycle until you’ve got a game that reliably generates a good experience. Once you have a game that does something good you can build the rest of your marketing strategy on top of that strong foundation. Let’s call this the “organic growth” strategy.
In this model, instead of assuming that playtesting is something that “happens naturally” to some variable percentage of the population when exposed to an in-development game, it treats it as a task which requires work, and it leaves it to the implementer to develop the strategies or tactics to get that work to happen (just like the previous model left it to the implementer to figure out how to get people to read the materials once they have them, or figure out how to get them to play the game once the materials are read). One “strategy” for getting playtesting to happen might be to have a cadre of supportive friends and acquaintances who will playtest your stuff for you because they like you as a person. Another might be to have employees who playtest as part of their job. Another might be to form a system of mutualism where designers participate in playtests of each others’ games for mutual benefit. (This last one is analogous to the way many writers develop novels: They have a writers’ group who read and give reactions to each others’ drafts, even if they might not be the exact target market of the novel). Finding effective strategies or tactics for this part of the model is hard!
But there are hard parts with the funnel model, too. For example, if you’re relying on “natural” playtesting you have to deal with the fact that many people don’t know how to playtest well (maybe they go into “how can I break this game?” mode when what you needed to know was “is this game fun when played normally?” and playing with someone who’s trying to break the game can contribute anti-fun for the other players). You also have to deal with the tricky process of separating useful feedback from distractions, or figuring out the alchemy that transforms feedback into progress on the game design (there’s an old chestnut about how people giving you feedback are almost always right about there being a problem but almost always wrong about what the solution is, but this is complicated by the fact that many people put on their game designer hats while playtesting and see “problems” that are hard to distinguish from “you haven’t implemented my preferred solution”). In fact, these sorts of difficulties have convinced some people who use the funnel model that playtesting itself is a largely useless process, except as a testbed for exploring a marketing strategy (this stance, naturally, is hard for the people using the organic growth model to wrap their heads around, because for them abandoning playtesting would mean bailing out before step one). Another issue with the funnel model is that discovering a success strategy at one level may provide constraints on the rest of the process. For example, a subsystem with a grabby gimmick can help convince people to play a game once they’ve read it, but if the success path of the game becomes contingent on that subsystem then any further game revisions will necessarily involve changing the rest of the game to work with that subsystem rather than the other way around.
In the independent tabletop RPG design communities that I’m aware of, most people seem to be operating from the assumptions of the funnel model. Personally I’m not sure that’s wise, since some of the level transitions seem to encourage strategies that can easily turn into wasteful zero-sum arms races (for example, if lots of people spam “please playtest my game!” messages on a forum then they’re competing with each other and also wearing out the patience of the forum’s readers). The big downside of the organic growth model is that the “get playtesting to happen” process looks to many people a lot like “and here a miracle occurs” in the current environment.
Let’s imagine a chess game. Two players who both know the rules sit on either side of a board with the appropriate pieces on it. To play, they’ll use their knowledge of how the pieces move, their mutual knowledge of the rules and victory conditions, the current position of each of the pieces on the board, and a mutually remembered bit of information about whose turn it is to make the next move. Obviously there are a few things that could mess this game up. A freak windstorm, for example, could blow all the pieces off the board. Or maybe a loud noise will distract the players for a moment and by the time they’re ready to return to the game their memory of whose turn it is won’t match because one of them (which one!?) got confused during the interruption. Or maybe one of them will do something that uses one of the more exotic rules, like en passant, and they’ll discover that their mutual understanding of the rules of chess isn’t as mutual as they initially thought.
Now let’s imagine that one of these chess players goes on an expedition to Antarctica but still wants to play chess with his cold-averse friend. They still can! What they decide to do is set up two different chess boards, one in Antarctica and one back home, and communicate their moves back and forth through whatever form of long-distance communication they can. When Antarctica-guy physically picks up a pawn on his chessboard and moves it to a different space he just tells his friend which pawn he moved where. The at-home friend picks up the corresponding pawn on his chessboard and moves it to the corresponding place to represent his friend’s move. All the rest of the stuff is the same: the important thing about chess isn’t that there’s a single physical board between the players, it’s that there’s an agreed-upon representation of the current game-state between the players. Having a single physical board certainly makes that easier and more convenient, but the important thing about the game isn’t how it’s physically implemented, it’s how it looks to the players. Each of the players can look at “the” chessboard and make their moves based on the current game-state. It doesn’t matter if “the” chessboard is a convenient fiction for two different physical chessboards that are being kept in synch by an extra process that isn’t normally necessary.
But what if these friends realize that they don’t really like chess that much and want to play something a little more action-oriented? They decide to switch to a first-person-shooter video game played over the net. Conceptually this isn’t too different from the long-distance chess game, but there are a few details that contribute some nuance. One important difference between chess and an FPS is that the turn-based nature of chess provides an easy interface-point for long-latency communications. If it takes much longer for one player’s move to get communicated it just looks to the other player like a really long turn. Since FPSs need to maintain a smooth, continuous-action flow of play they need to have the effect of the moves represented immediately. When the Antarctic player presses his “shoot” button he’d better see his character start shooting right now! The two computers are both running instances of the game, but the other guy’s doesn’t realize the first started shooting until a message dispatched over the network gets to his computer. But maybe at the same time that the Antarctica guy decided to shoot his gun, his target pressed his “run” button and started moving. In Antarctica, the player thinks he’s shooting (right now!) at a stationary target, but at his friend’s house he thinks he’s moving (right now!) and not being shot at. From the Antarctica perspective the shot should hit (assuming the aim is good) and the target should be wounded. From the other perspective he shouldn’t be wounded at all: nobody was shooting, and even if they were his character wasn’t at the place that the bullets would hit! The two simulations aren’t perfectly consistent. But they don’t have to be! As long as they’re close enough, the players won’t notice. As a human player, the warm guy doesn’t know with perfect certainty where the Antarctica shot was aimed, so if the game has an under-the-hood mechanism that gives “hit detection” precedence to the shooter’s POV then the Antarctica computer can tell the other one not only that a shot was fired but that it hit. The at-home computer can play its “gunshot” sound effect, display the “shooting” animation for the other character, reduces the hit points of the target, and most of the time it will seem perfectly normal to the at-home player that the other character shot and hit him at his current location. The important thing to notice is that there doesn’t need to be a single authoritative game-state in a single place in order for both players to feel like they’re playing the same game with the same state. As long as it looks close enough they won’t realize that their two computers are not exactly on the same page at every instant.
As players they maintain the convenient fiction that they are in the same world because the “game” involves making decisions as if you were, just like it makes more sense to interpret what they see on their screens as a window into a 3D world rather than a bunch of pixels on a flat display. Just like it’s not useful when playing long-distance chess for them to dwell on the fact that they don’t have a single physical board between them, it’s not useful for them to dwell on the potential artifacts of network gaming (unless the distortions become so extreme that they overwhelm the suspension of disbelief and they have to give up because there’s “too much lag” over their network). By buying into the illusion of consistency between the somewhat-independent computers they can play this type of game together.
Now let’s imagine that the adventurous friend returns from Antarctica and the two of them get together to play another kind of game they enjoy, a tabletop RPG. Here they also need to maintain a sufficiently-synchronized game-state in order to play. To do so, they buy into the convenient illusion that there’s a single “fiction” or “Shared Imagined Space” between them. They probably have some concrete common physical touchstones like dice or character sheets as part of the game, but a big part of play involves their brains independently keeping track of the current game-state of imaginary people doing imaginary things, and they send messages back and forth to keep each other more-or-less in-synch (using high-tech “talking” technology). Since their brains aren’t as simple as chessboards they can’t rely on being perfectly in-synch at all times, so their game needs to be constructed in a way that encourages and eases synchronization on important points. For example, if their game has a mechanic which gives a “high ground” advantage then the players will be primed to pay special attention to character altitudes relative to each other in “the” imaginary world. Maybe their mental picture of the characters won’t agree on points like whether or not they have mustaches, but they are likely to agree on who is higher than who if they both believe that is important to the game.
Being sufficiently synchronized to game is the foundation for a functioning RPG (and the astute reader will notice how weaselly a word “sufficiently” is). Many RPG techniques and design elements serve to maintain that synchronization. For example, the “fictional trigger” in an Apocalypse World move can serve like the snap-to-grid functionality of a computer painting program to snap the “fuzzy” mental images of the different players around easily-communicated concrete templates. If my character seems close to “Going Aggro” on somebody I am pulled toward embodying that in my roleplaying because I know that the other players are watching for whether characters are Going Aggro and will understand what I’m thinking better and be more easily synchronized to what I’m imagining when they can use that concrete and mutually-understood pattern as a touchstone for how the scene should be playing out in their imaginations. Agreeing with the other players that the “Go Aggro” move should be invoked and starting the corresponding mechanical procedure gives us an explicit way to acknowledge synch-points without drawing unpleasant attention to our efforts to keep synchronized.
I’m not an expert on Topology, but one of the ways I think about the games I like is that they make use of the idea-space inside the human brain as a gameable space. Now, by that I don’t mean that you can imagine places that aren’t real and think up activities people might engage in in places like that. What I mean is that the way we think actually provides “dimensions” along which you can design meaningful interactions in a game. From my reading of what contemporary psychology and cognitive science tell us, we’re capable of perceiving the appropriateness or congruence of matches between ideas. You know the confident “that feels right!” feeling you get when you figure out what the answer to a riddle must be, or when you come up with the perfectly apt humorous remark? Or the “that’s not right” feeling you get when Hollywood miscasts a part in a movie adaptation of a story you know? That’s what it subjectively feels like to have different levels of connection between ideas, which is apparently how the intuitive side of our cognition works. Even though we don’t have scientific units for it, we can get a feel for how “librarian-y” someone is by intuitively comparing them against the idea of “librarian” we have in our brains. We can even get a feel for how weaselly someone is.
This is the entire basis of the game Apples to Apples. In it, one player puts an “adjective” card on the table (for example, “Hot and Sticky”). Then all the other players consult their hand of “noun” cards and put forward the one they think the initial player will select as the “best match” (for example, “The Equator”, “Cinnamon Buns”, “The Sports Illustrated Swimsuit Issue”). Some people have a hard time grasping that there is no single way that people are required to make that “best match” comparison. It’s not always “the most similar” or “the most opposite”. Sometimes it’s the ill-defined “funniest”, but even inveterate jokesters will sometimes feel compelled to pick a straightforward match if it’s dead-on. The way it works is that the player compares the “adjective” they put forward to the various options and picks the one that feels like the “best match”. We don’t need to put a name to a comparison to feel how strong it is. Strictly speaking Apples to Apples tends to be about emphasizing the minor variations between people rather than the commonality because it asks the player to pick a “best” match on each round (thus the way to win, if you care about that, is to “play the player” and put forward cards with matches that are likely to resonate especially strongly), but it illustrates the point that there are dimensions of play that games can lean beyond simple factors like tallest/short, fast/slow, near/far, big/small, etc. Personally I’m not a huge fan of the gameplay in Apples to Apples (my sense of humor tends to run a little more cerebral and surrealistic than average so my joke answers nearly always lose out to the more obvious jokes) but since it uses this abstraction as the central element of play it’s a useful example.
While they don’t always foreground it the way Apples to Apples does, Roleplaying games make heavy use of this concept to inform and constrain play. The old-school “puzzle solving realism” style of play, for example, leans heavily on the ability of humans to mutually imagine “that’s what would happen!” to explore the consequences of poking things with ten-foot poles or pouring acid on them. The Burning Wheel family of games orients players to judge characters by looking through the lens of written character Beliefs, rewarding players for acting along (or dramatically against) the line of those Beliefs. Games with oracle mechanics like Ganakagok use abstract concepts to guide play (“figure out the most ‘Woman of Storms’ way to conclude this scene”). Even something as fuzzy as “what’s the most dramatically appropriate (or dramatically ironic) thing?” or the dreaded “what’s best for The Story?” can be used in a game context. Stories and storytelling have a huge role in human culture and the way that human minds work, so it shouldn’t be surprising that we have a lot of intuitions related to stories and imagination. These intuitions can be built into the “space” of play in these games in the same way that features of human locomotion are as important dimension of play in sports as ball-physics and field geometry.
When analyzing systems that operate on information it’s often valuable to consider how that information matters to the control flow of the system, and games definitely fall into this category of system. One big distinction between types of information is discrete vs. continuous, or digital vs. analog. A continuous “variable” can be any value within a range: think of something like temperature, distance, or time. A discrete variable can only be in one of several mutually exclusive states: on/off, in-bounds/out-of-bounds, too-big/too-small/just-right, etc. Continuous variables are really useful because that’s how almost everything in the actual world we live in works. Discrete variables are really useful because it’s possible to build simple procedures around them: if A do X, but if B do Y.
As a simple but nontrivial example think of a thermostat. It has three continuous inputs: the current temperature, the low set-point and the high set-point. The thermostat is in charge of the heater and knows and controls whether it’s currently on or off. Internally it doesn’t really do anything with the temperature directly, it uses a comparison to create a discrete variable from two of it’s continuous ones: “is it currently hotter than the high set-point?” and “is it currently colder than the low set-point?”. Operating on these discrete concepts lets it make a decision that’s simple enough for it to apply to the binary world of “should the heater be burning right now?”: if it’s hotter than the high set-point and the heater is on, turn it off, but if it’s colder than the low set-point and the heater is off, turn it on.
Lots of games have things like this, too. In soccer, the ball is somewhere in the three-dimensional space where the game is being played, and this feeds into discrete categorical concepts like “is the ball currently in-bounds?” that are used by the game procedures to control the flow of play. In baseball, whether a pitch counts as a “ball” or “strike” corresponds to where it travels through the strike-zone of the batter. In the UFC mixed-martial-arts organization some moves are legal and others, such as punches to the back of the opponent’s head, are illegal. When you look at these distinctions from the digital side of the analog/digital divide there are obvious and categorical differences between them: the difference between an in-bounds ball and an out-of-bounds ball are night and day! From the analog side it can be fuzzier: what if the ball is right on the edge of the line? What about a pitch that’s just grazing the edge of the strike-zone? Heads are kind of round, so the distinction between side and back is not always obvious, right?
In games, translating from the continuous/analog domain to the discrete/digital domain of the rules and procedures of the game usually involves human interpretation or judgment. Oftentimes games will give one participant, such as a referee, a special privilege of having authoritative judgments or interpretations, but even in games like that all of the participants need to understand how those interpretations and judgments will be made and make their own. Soccer players don’t want to play on a field where the lines are invisible to everybody but the refs, they need to be able to predict the rules-consequences of their interactions with the ball in order to play. They may not be able to exactly predict how the ref will make the call in edge-cases, but they can reasonably expect that their own interpretation will be similar to the “official” interpretation, so they can use their own interpretation as a good proxy for evaluating what kind of move they want to make in the game. (And plenty of casual sports are played without an officially designated ref, the players just use some other process, sometimes ad hoc, to resolve edge-cases if there’s no widespread consensus interpretation). Similarly, the intention of the “no strikes to the back of the head” rule in the UFC isn’t to give penalty points to inaccurate punchers but to discourage fighters from engaging in behavior that the UFC has decided is too dangerous: the ref makes the authoritative call in the octagon, but the most important impact of the rule is on the fighter when he decides whether or not to throw a punch based on where he thinks his opponent’s head will be when the punch lands.
Many RPG rules operate on things happening in the analog world of “the fiction” so they have lots of these interpretation elements cooked into them, so looking at the nuances of these interpretative processes is obviously very important in RPG Theory. But we shouldn’t mistake the importance of this concept to RPGs for the idea that interpreting or translating from continuous to discrete concepts is something unique to RPGs. The interplay between the interpretations and judgments of different participants in an RPG is an interesting and important topic if you’re trying to understand RPGs. The interplay between the interpretations and judgments of different participants in a pitcher/batter interaction is an interesting and important topic if you’re trying to understand that part of a baseball game.
(Also, I’ve tried to use simple examples in this blog post in order to write with clarity, not to deny the existence of subtlety. My claim here is that both “is that really Go Aggro?” and “is the ball really in-bounds?” are both examples of interpretation that feed into rules. It can be easy to get distracted by the simple one-dimensionality of the in-bounds/out-of-bounds thing because we can easily imagine constructing a simple mechanical or electronic device that we could rely on for official in-bounds/out-of-bounds rulings while the only thing currently known that can do the Go Aggro thing is a human brain. That’s an important difference worth thinking and talking about! But it’s also worth realizing that “how hard would it be to build a robot referee?” is a different question from “how are the players interacting with this game?”.)