The Delta Theory of Meaningfulness

[NichG]

@NichG - Where to start?

For the Cunning Strategist, it's not goalpost moving, it's not missing the point, it's... a really odd miscommunication? So, for our "spherical cows in a frictionless vacuum", let's assume in a conversation about Agency that there is no "noise", and all... Agents(?) successfully achieved all Goals that were valid / possible in the scenario. Does that... no, that doesn't work with what I said, either. Sigh. I see that, juxtaposed with the "child romance" example, one might assume that "Cunning Strategist" was a joke, but I actually meant it seriously, as in "someone who achieves their goals through unconventional and unpredictable means" (obviously that's not a definition of that term, simply an explanation of what I meant to imply in that context, rather than the opposite of a "cunning" strategist, whose actions are unpredictable because they're (unsuccessful and) dumb).

Anyway, that aside, what are our knowns and unknowns in this... example? conversation? theory?... you are describing? My limited experience in this field involves the company starting with only data, attempting to locate patterns, attempting to hypothesize names (ie, "goals") for the buckets of patterns that were created, manipulating the environment (in this case, the code) to attempt to facilitate these hypothetical goals (or to facilitate the company's goals, given said hypothetical customer goals), and observing the results. In other words, the worst, dumbest blind Business Analyst through Data Analysis I can imagine. Is my roundabout way of saying, warning: I'm a bad audience, because my experiences have left me biased against this field.

ANYway, what I was asking was, with what you were describing, what were the inputs, and what were the outputs? Are you walking in with a finite set of modal goals, and trying to parse the data for patterns that allow you to sort users into modal buckets? Because it sounded to me like one of the outputs you described was "these are the decisions that matter for differentiating between goals", but I was uncertain what you were claiming were the required inputs in order to achieve that output.

You know, I think I break character more when talking with you than with anyone else.

What makes this particularly complex is that, with sufficient X (am I committing the Fallacy of Four Parts or otherwise obfuscating differentiatable details if I call X "Agency"?) there are multiple paths to goals; someone with goal A may usually take action A*, and someone with goal B may usually take goal B*, but someone with goal B could take action A* when taking a different path to B. Or someone with the goal not!A could take action A* simply to obfuscate their motives if their goal was "not to be detected" (someone with an unpopular goal, a spy, someone trying not to show favoritism, or my favorite half-remembered example where a girl points out to a (depressed?) woman that her love interest talks to everyone but her -> he's in love with her too and trying to hide it. And, yes, there is so much data in each choice (I'm reminded of the 12 spheres puzzle), compounded by the fact that "becoming king" could be a goal for one person / character, but a side-effect for another's choices.

So, yeah, with all that said, I'm struggling to ascertain just what modern data theory can tell us in this setup. But I do like the way you framed it in terms of Expression - any elaboration along those lines would also be welcome.

The 'spherical cow' bit isn't the absence of noise, but rather the presence of omniscience. For the formal definitions the mathematical framework doesn't assume a limited viewer within the world observing the agent, it takes a god's eye view in which basically know and tabulate everything. Every possible trajectory of every possible universe in which every possible action and every possible outcome of all sources of noise are taken. So for the math, there's no such thing as a divide between conventional and unconventional. If the strategist's actions were weird but increased the chance of them achieving state A* by 5%, then that's exactly the same as someone who took boring, common sense actions and also increased the chance of achieving state A* by 5%.

That's kind of the point of the math, to give us a way to remove our own subjective preconceptions. Some of the applications for this kind of framework are like, lets say we encounter Star Trek style alien life like sentient solar flares or mycorhyzhal crystals growing in the crust of a plant or entities that exist by manipulating quantum randomness or something - we want a mathematical framework that, at least asymptotically given sufficient observations and information, would tell us 'yeah, that vibrating rock over there is an agent, and its trying to create aesthetically pleasant triangles in the shadows of trees that will grow here in 1000 years by vibrating in such a way to encourage or discourage plant growth around it by manipulating where the roots go'. Or like, take some really huge Conway's Game of Life run on a 10^12 by 10^12 grid and highlight all of the 'agents' without a human having to look over everything and make the judgment by hand (which is likely to miss stuff anyhow, because of preconceptions).

The cost of that kind of generality is that you have to know everything that could be.

But, if you want to use this for practical (but still quantitative, not just conceptually inspirational) purposes, you do what you always do with these uncomputable mathematical frameworks - you approximate them. In which case, they become computable but sometimes wrong. In practice, this thing is built out of various kinds of mutual information, and you can estimate or at least bound mutual information pretty well with neural networks (when you train a network to output probabilities the difference in the training loss between start and end is a lower bound on the mutual information between input and output). In which case, yeah, an 'unconventional' strategy may land in a generalization gap of one of the networks and get missed. But then that's the fault of the approximation method rather than a fundamental issue, and it can be detected after the fact (hey, the network thinks this guy shouldn't be reaching A*, but he keeps reaching A*!) and even includes some of the seeds of its own repair (okay, lets train the network on this guy's runs, then in the future the network will be able to anticipate this kind of strategy).

Goals too can be quite general in this kind of framework - the most general case is that the space of goals is the space of all possible reward functions over states rather than lists of things a reasonable person might want to achieve. You can impose a prior or crunch this down - the cost of doing so is you might miss something, the benefit being that you turn the intractable computation a little more tractable. Maybe even introduce some interpretability!

So like, if you wanted to use the whole framework quantitatively and in practice, an example might be something like iteratively improving the design of a game like Minecraft by collecting lots of beta-tester playthroughs, finding the particular choices that have high mutual information with what the beta-tester eventually did, then tweaking the design of things to make those choices exist more evenly within the course of play, or hit a certain pattern of expanding opportunities you want as a designer, or maybe it points you to 'obligate' subsequences in a crafting system that aren't particularly interesting (lots of Minecraft mods suffer from this, with long crafting sequences that have no internal branch points - general resource A -> intermediate B -> intermediate C -> final product D, but B and C only ever make D).

[Mordar]

I think we all have said, in some form or fashion, that "agency" or "impact on the events of the game" or whichever phrase we prefer, is one quality that can influence the "goodness" of the game. NichG suggests it is a multiplier, I think it is a mid-impact slider, others may find it lower or higher on the spectrum than either of us. Some players might even find it a complete non-issue.

I think a number of other factors make games meaningful, and probably only half of them are related to the actual game component. If we stick to just things in the game (so not related to the social dynamics, etc) I think compelling questions, compelling settings, compelling ancillary characters, probably in that order, are terribly important. Then, depending on those elements, my ability to shake the pillars of heaven can be considered...a game wherein I cannot effect broad change isn't de facto a bad game for me, just like a game where I can reshape the entirety of the known world is not necessarily a good game.

I believe in the social contract of RPGs, the shared experience, and the commitment of time and effort to one another. Trust matters, but I am willing to actively choose to accept boundaries for a specific game until such time as the game convinces me it was a bad idea.

I expect my player skill, character skill (inclusive of all elements...levels, feats, whatever) and circumstance to influence my likelihood of success in actions, both narrow/immediate (like climbing a wall), broad/intermediate (like infiltrating a mob), or expansive/extended (like becoming the Archmage of the Order of Flightless Birds). I expect that GMs will not adjudicate actions based on their wanting to "beat" the PCs. I accept that Quantum Ogres can regularly improve games.

Finally, I think there is a risk people misconstrue failure with lack of agency. If the choices I make don't have the outcome I want that doesn't mean that I didn't influence that outcome. If I made a character build that excels in ActionA (so I have a +10 modifier), I make equipment and deployment decisions to improve my chances in ActionA (giving another +5 modifier), I might still fail at ActionA especially in an isolated instance...because I rolled a "2", because it is the hardest ActionA instance I have ever faced, or because something was actively working against my ActionA. But I had a helluva lot better chance that Quertus because that Mage can't ActionA to save his bacon. I influenced the chances of achieving the outcome, but still failed. Same thing applies to a broader goal, but it is less easily seen, because I am going to ActionA successfully often enough to recognize my influence on the outcome, but that whole achieving Archmage of the Order of Flightless Birds things can only be addressed once.

tl;dr: Want to make a game more meaningful to me as a player? Give me a compelling setting with interesting questions and some good characters. As a player I want my GM to give me a great framework and I'll fill details.

- M

[Quertus]

So like, if you wanted to use the whole framework quantitatively and in practice, an example might be something like iteratively improving the design of a game like Minecraft by collecting lots of beta-tester playthroughs, finding the particular choices that have high mutual information with what the beta-tester eventually did, then tweaking the design of things to make those choices exist more evenly within the course of play, or hit a certain pattern of expanding opportunities you want as a designer, or maybe it points you to 'obligate' subsequences in a crafting system that aren't particularly interesting (lots of Minecraft mods suffer from this, with long crafting sequences that have no internal branch points - general resource A -> intermediate B -> intermediate C -> final product D, but B and C only ever make D).

I'm just gonna poke at this one piece. So, it sounds to me like the A->B->C->D chain was supposed to represent no significant Agency, but I have a question: What if A, B, C, and D are ends in and of themselves, not just means? Even if that's the only possible build structure (or the only build choices ever made), I feel there's Agency in choosing to use B, vs crafting it into C. Like if Mithral Ore can be smelted into Mithral Bars, which are turned into Mithral Rings, which you can used to craft Mithral Chain... and people can place Mithral Ore blocks to build stuff, and stack Mithral bars to make pretty displays, and wear Mithral Rings as jewelry (OK, Minecraft doesn't really support this) in addition to making Mithral Chain, and people do all of these, that sounds like a choice. Or are you saying that those would be represented differently, as [A, A->B, A->B->C, A->B->C->D], rather than all lumped together as A->B->C->D?

If we stick to just things in the game (so not related to the social dynamics, etc) I think compelling questions, compelling settings, compelling ancillary characters, probably in that order, are terribly important. Then, depending on those elements, my ability to shake the pillars of heaven can be considered...a game wherein I cannot effect broad change isn't de facto a bad game for me, just like a game where I can reshape the entirety of the known world is not necessarily a good game.

I think I have to agree with you. The only point of contention, the only potential difference in our preferences might be related to just how much Agency is inherent in the concept of having "compelling questions", and maybe whether those compelling questions can be of the form, "here is the situation, what do you do?" rather than more "traditional question form" (like, "what is one's moral obligation upon seeing a wounded person along the side of a road?" or "what are the criteria for Personhood?"). And maybe just a hair of balancing the number of yes/no binary questions vs open-ended questions.

I accept that Quantum Ogres can regularly improve games.

Rather than me give a lengthy spiel, let me start with, what do you mean by this / why do you believe this?

Finally, I think there is a risk people misconstrue failure with lack of agency. If the choices I make don't have the outcome I want that doesn't mean that I didn't influence that outcome. If I made a character build that excels in ActionA (so I have a +10 modifier), I make equipment and deployment decisions to improve my chances in ActionA (giving another +5 modifier), I might still fail at ActionA especially in an isolated instance...because I rolled a "2", because it is the hardest ActionA instance I have ever faced, or because something was actively working against my ActionA. But I had a helluva lot better chance that Quertus because that Mage can't ActionA to save his bacon. I influenced the chances of achieving the outcome, but still failed. Same thing applies to a broader goal, but it is less easily seen, because I am going to ActionA successfully often enough to recognize my influence on the outcome, but that whole achieving Archmage of the Order of Flightless Birds things can only be addressed once.

Hahaha, I almost missed this bit, because I figured I knew where you were going. I guess... hmmm... I prefer that Expression be dominant whenever possible in such a scenario, in a "Kiss the Girls" way, such that you'd get the opportunity to succeed at ActionA (your "hardest instance I have ever faced" and "often enough") to demonstrate that facet of the character before having the chance at failing ActionA. And it's great Expression when you can also succeed in comparison to the Expression of someone (like a signature academia mage) who just can't even ActionA.

tl;dr: Want to make a game more meaningful to me as a player? Give me a compelling setting with interesting questions and some good characters. As a player I want my GM to give me a great framework and I'll fill details.

- M

I always felt the question of Agency was often curtailed when, rather than let you "fill details", the GM rejected anything but what they wanted in those blanks.

[NichG]

I'm just gonna poke at this one piece. So, it sounds to me like the A->B->C->D chain was supposed to represent no significant Agency, but I have a question: What if A, B, C, and D are ends in and of themselves, not just means? Even if that's the only possible build structure (or the only build choices ever made), I feel there's Agency in choosing to use B, vs crafting it into C. Like if Mithral Ore can be smelted into Mithral Bars, which are turned into Mithral Rings, which you can used to craft Mithral Chain... and people can place Mithral Ore blocks to build stuff, and stack Mithral bars to make pretty displays, and wear Mithral Rings as jewelry (OK, Minecraft doesn't really support this) in addition to making Mithral Chain, and people do all of these, that sounds like a choice. Or are you saying that those would be represented differently, as [A, A->B, A->B->C, A->B->C->D], rather than all lumped together as A->B->C->D?

If they're ends in themselves that's one thing, but in the case of the actual examples I'm thinking of (from mods like Immersive Engineering and the like) there are lots of these things that are just inventory entities that you can't even place in the world. So like, once you've made a 'bucket of creosote' or 'a pile of iron dust', it doesn't actually do anything until you do two specific following steps. You have to wash the iron dust to get refined iron dust, then melt the refined iron dust into iron nuggets, then smelt the iron nuggets into iron bars. There's no branch points and the intermediates don't show up in any other crafting recipes and can't be placed in the world, so you've basically said 'my goal is to have iron bars' and the mod just expands that production pathway into extra steps. That first choice to make the iron dust is the only agentic choice in that process (because otherwise maybe you could have placed the iron ore as a decoration, or smelted it directly, or something like that).

There are agentic choices enabled by the existence of that process, namely, do you build all of those machines to get 3 iron bars from 1 ore, or do you just use the base game smelter to get 1 for 1? But that choice happens once, whereas this production pathway you're doing over and over (though to be fair, the kinds of mods that have these things are intended to all be automated).

So this way you could see that 'ah, the choice to make iron dust tells me about the player and their context, but the choice to wash the iron dust doesn't actually provide me any additional information. Maybe I can add some other recipe that uses the iron dust to go some other direction, like alloy it with coal dust to make steel?'

Actually now that I think about it, at least one of this family of mods does let you diverge iron dust into a recipe for steel, so huzzah for that! There are a lot of these kinds of mods and honestly they all blend together for me by now (last time I played modded Minecraft, there were like three separate 'copper ore' entities that you had to get a separate mod to make them interchangeable...).

[gbaji]

This is why I'm pushing against that definition of agency, because it really makes things about 'what things must the GM not do?' versus 'what are some things the GM could do, that would help people make a greater connection to the game?' If you think in terms of norms, expectations, etc, then all that lets you talk about is the errors. 'Violations', 'curtailing', 'what is acceptable', etc. But I think that's a bad way to think, its inherently a defensive position where you're looking to excuse things you're doing as part of your job to people who might criticize it. Its an easy stance to fall into on a message board like this, because its basically the first thing that's going to happen if you say you did anything. But ultimately I think its unhelpful.

I don't think that's necessariy about the definition of agency though. It's about the fact that agency is not the sole determinant of "meaningfulness", which was the original point of this thread.

It does seem that what you are talking about now is how the GM can inject that meaningfulness into the game. And IMO, that has very little to do with player agency, and a whole lot to do with the GM being creative and placing interesting things to do, places to go, people to interact with, evil schemes to overcome, objectives to strive for, etc into the game, and then presenting these things to those players in a way which is both engaging and enjoyable for all involved.

But yeah. Agency is an entire different kettle of fish. I suppose it does intersect at the point of the GM, in the pursuit of said meaningfulness in the game, potentially restricting/curtailing player agency. If the GM decides "it would be really fun if .... <some sequence of events occur>", and the GM falls into the trap of trying to force that "interesting sequence" to happen, even if the players are trying to go in a different direction, or generate their own alternative sequence of events, this can result in a problem in the game. So I suppose one could say that the trick for the GM is how to put that meaningfulness into the game, without it becoming the GM writing a script and then the players just playing their parts in that "meaningful story".

It would be better, IMO, to say 'what are some neat things I can add to a game, and how can I know which ones are likely to work better?'. Rather than seeking justification for the game you were going to run anyhow, the focus is more on having more tools to run the game maybe you wish you could. So like, instead of looking at a situation where a player strongly asks to achieve something unreasonable, look at the situation where you have a player who is playing somewhat passively and reactively. Maybe you don't directly know what it is they want at all, what sort of things can you add to your game that will give you a higher chance of catching their attention, getting them excited and participating actively, and caring about the outcomes of the campaign?

Absolutely. And I also think, following on to what I wrote above, that there is no single perfect answer for this. As you point out, some players are very passive and are waiting for the GM to give them things to do. Others are very active and will find things to do. The GM needs to recognize each type of player and respond accordingly. I also tend to think that the passive player may be the hardest to engage while not curtailing "actual player agency", and honestly probably the hardest to even detect as well. When a player is very passive, you don't always know what they are trying to do themselves, and it can be very easy to, in the pursuit of providing them motivation and "meaningfulness", not even realize that you may be scripting things for them a bit. I suppose one suggestion is to talk to the player more directly and try to get an idea of what sorts of things they want to do or achieve in the game.

Proactive players are, as one might expect, the exact opposite. They are the most likely to push for things that are actually unreasonable/unrealistic and thus may generate the most claims of their agency being curtailed (more likely to push against the boundaries). However, they're also the most likely to generate idea that assist the GM in "creating meaningfulness" in the game. So usually a pretty decent tradeoff. It also takes a bit of balance to manage this (just a different kind of balance).

Again though, I really do think that the starting point is for the GM to first establish a set of standards and "norms" for how the game setting works. This gives the GM tools to determine when the baseline is being pushed too far, both in their own actions and in those of the players. If you don't have this in your setting, the you wont know if whatever "new and intersting thing" you are doing is actually impacting agency along the way. Again, this of course assumes that the objective here is to figure out how to add that meaningfullness to the game while also enabling maximum possible agency for the players. Obviously, if for some reason you actually want to run a game that is more in one direction or the other, then there are different measurements to use.

I still think it's useful to have that starting point though. Once you have that, you can measure everything you and your players do in your game against it.

[Mordar]

Rather than me give a lengthy spiel, let me start with, what do you mean by this / why do you believe this?

[Using the idea of a planned encounter (an ogre attack) that the GM deploys in a location other than where they originally planned...or that they had not fixed to a location = Quantum Ogre]

I am at the game to participate in the traditional game, and I want to give my character a chance to interact with the world. I want to interact with the encounters that my GM created to be part of an entertaining and enjoyable game. If the reason we did not interact with that encounter was simply because we chose to go to the Village of Hamletsville instead of the Hamlet Villville, and there is no compelling reason the Ogre could only be on path to Villville, then I have missed out on a fun thing.

That being said, I do not view it the same way if we are intentionally going to the Village of Hamletsville because we want to avoid the rampaging Hamlet Villville Ogre bandit that we heard about in town...that usually (not always) will be an irritation. Unless, of course, the Quantum Ogre had a specific reason to be on the lookout for our party and had reason to know where we were going...and it darn well better be made clear in the encounter that this is the case.

In short: The GM that I am choosing to spend time playing this game with has chosen to spend time crafting an encounter, and barring any compelling reason to not have the encounter, I want to have the encounter and be entertained.

Note: Encounter can be non-combat, can be discovery, can be a bijillion things.

Hahaha, I almost missed this bit, because I figured I knew where you were going. I guess... hmmm... I prefer that Expression be dominant whenever possible in such a scenario, in a "Kiss the Girls" way, such that you'd get the opportunity to succeed at ActionA (your "hardest instance I have ever faced" and "often enough") to demonstrate that facet of the character before having the chance at failing ActionA. And it's great Expression when you can also succeed in comparison to the Expression of someone (like a signature academia mage) who just can't even ActionA.

Word.

Strange example, not sure what it means exactly, but know we hated it, and it seems to relate:

Spoiler: Negating other characters

Show

Playing 7th Sea, but would apply to many of the non-D&D games out there. Player designed a character to be really good at a few academic exploration kinds of skills that require active learning (don't remember exactly which, but it was like Cartography, History, stuff like that). Another character was a street-rat who "grew up near the University" and was built as a combat character but with like 5 dots in Wits and a GM-fiat allowance to roll on skills that require ranks "...because zero ranks is ranks". So the girl playing the academic had the chance to shine and rolled adequately on this important thing to know to continue the adventure...and the munchkin promptly declared they were rolling as well and got a significantly better result (dice luck plus some broken concept advantage). That was the point at which the girl playing the academic utterly disengaged from the game. Combination of niche-stepping and inadvertent negation of the entire character concept. The combination of decisions by a well-intentioned GM and another player zeroed out the meaningfulness of the game.

- M

[NichG]

I don't think that's necessariy about the definition of agency though. It's about the fact that agency is not the sole determinant of "meaningfulness", which was the original point of this thread.

It does seem that what you are talking about now is how the GM can inject that meaningfulness into the game. And IMO, that has very little to do with player agency, and a whole lot to do with the GM being creative and placing interesting things to do, places to go, people to interact with, evil schemes to overcome, objectives to strive for, etc into the game, and then presenting these things to those players in a way which is both engaging and enjoyable for all involved.

But yeah. Agency is an entire different kettle of fish. I suppose it does intersect at the point of the GM, in the pursuit of said meaningfulness in the game, potentially restricting/curtailing player agency. If the GM decides "it would be really fun if .... <some sequence of events occur>", and the GM falls into the trap of trying to force that "interesting sequence" to happen, even if the players are trying to go in a different direction, or generate their own alternative sequence of events, this can result in a problem in the game. So I suppose one could say that the trick for the GM is how to put that meaningfulness into the game, without it becoming the GM writing a script and then the players just playing their parts in that "meaningful story".

The original post of this thread resonated with me because I think while what you're saying is sort of a classic view, those things actually don't matter to me *except* in how they give me things I can explore via acting through them. And it explains campaigns I've found evocative or boring a lot better than e.g. interesting setting or creative GM. Its not that GM creativity doesn't factor in, but only the parts that make it through the filter of agency actually matter to me. I've had very creative GMs whose creativity is primarily 'on their side of the table', whether that made sense or not in the norms of the setting they presented, those campaigns didn't catch me nearly as much as the campaigns where the GM e.g. gave the players opportunities to play with things you normally wouldn't get to play with.

So this is why I say that agency (or informed delta, in the specific context of this thread) may best be seen as a multiplier. If you have a boring world and a lot of agency, maybe you're multiplying a lot of agency with zero things worth actually caring about - if you had goals, you could pursue them, but the world doesn't support interesting goals. On the other hand, if you have a really creative, evocative, etc, etc world and no agency, you're also multiplying by zero - the world might be interesting but its not actually sharing that interesting-ness with you in ways that can matter to you. There could be all sorts of goals worth pursuing but since you have no power to pursue them, they don't matter.

But the 'interestingness' of the world independent of the players relationship to the world itself is not, to me, indicative of what I will find to be a good generator of meaning. For me, justification within the context of the world, established norms, etc basically don't matter that much in of themselves. If a GM presents a very interesting world with lots of well-established justifications of why I can't do any of the interesting things in it, that doesn't make it suddenly like 'oh okay, I'm fine with having no agency' - it just means that entire framework contains very little to interest me, so from experience I should probably leave the game.

Its always harder to talk about my own GM-ing here because we can't usually see our own blinders, but of course I wouldn't be pushing this viewpoint if it were inconsistent with my own experiences as GM. I've run a lot of very experimental campaigns, and the things that tend to fail are predominantly when the pathway to connecting with the oddities or interesting or novel aspects isn't well enough marked. Thats why my current philosophy of system design and GM-ing is that rules are tools to give players OOC information to let them know that things are possible and reliable, which their preconceptions would tell them not to expect to work. You don't write a rule because 'we need rules to know how the game is played', you write a rule because its an efficient way to let a player know 'if you do X, I promise that Y will happen in such and such a way' because that makes 'X' available as an action with a predictable consequence, which in turn allows a possibility to turn into actual agency, which in turn means that players have that additional avenue to connect to things.