## Filtering

16 March, 2008 at 10:19 am (rpg theory) (, , , , )

Given that choice is and important part of roleplaying, how do people make choices? There are useful theories out there (including game theory, with which I have a passing familiarity). I’ll try building another one so that it hopefully has something useful to say about roleplaying.

This model is specifically about how people make choices in the context of roleplay and that affect the rp. Extending the model is not particularly hard and is left as an exercise for the reader. The limitation is essentially arbitrary.

I also try out sketching a model for immersion as it is seen through this piece of theory. The threefold model is another example.

### Some definitions

Filter is a criterion that assigns a weight to different options people have regarding their play. In case someone in the audience is mathematically inclined, a given filter is a mapping from the decision space (this is the emulation space [emspace] for those familiar with Kuma’s AGE model) to real line; generally speaking, the entire real line is not needed and one can work with a given interval or other subset.

Example filters: What would my character do? Will this result in total party kill? What will this say about me as a person? Is this appropriate to the genre?
Example filters for the more elaborate model: How much will everyone enjoy this? How genre-appropriate is this?

A filter is strong when it makes sharp contrasts between different options; likewise, a weak filter makes little difference. These are not exact definitions and should generally be used as a part of comparative phrases (they imply comparison to some unstated standard if not used in such a way).

## A simple model

Every participant has a number of filters. Each filter accepts certain options and rejects others. In math: Filter is an indicator function of a subset of the decision space. It assigns value 1 to any option that is within the subset and 0 to any that is not.

When making a decision, participant applies filters, one at a time, with each application further restricting the possible choices (or at least not adding any). When the remaining options are sufficiently close to each other, the participant makes that choice. “Sufficiently” depends on how important the occasion is, how tired the person is, level of attention, and so forth.

The order filters are applied in is a matter of playing style and other influences.

### Assumptions of this simple model

This is not a good model. It makes the following assumption (and others): That people always either reject or accept a given option; essentially, this model assume two-valued logic. This is not very accurate.

## A more elaborate model

Filters work as above, expect that they can get values up to an arbitrary positive value on the real line (this can, but need not, be fixed). That is: A given filter assignes some value to all potential options the player has at a decision point (in nontrivial cases the assigned values differ from one option to the next). For desirable (according to that criterion) options, the value is high. For undesirable ones it is low. These are multiplied over all the used filters until one is sufficiently larger than the others. The option with highest value is then implemented and a choice is made.

## Applications

Character immersion happens when player identifies with a character to great degree. Filter theory sees immersion as a process where the player has one or few very strong filters, so that it seems the player is not making choices at all. Typically this filter is “What does (would) my character do?”. This is clearly distinct from the disruptive behaviour sometimes known as “my guy-syndrome”, in which the relevant players uses the character’s actions as excuses for bad behaviour. In this model, players who disrupt the game and use the character as an excuse don’t make heavy use of the “what would my character do”-filter, but some other, more harmful, one.

The threefold model is a theory about filters. G, D and S all signify emphasis on given (family of) filters: dramatist GM has a strong “does this make a good story”-filter, for example. This is very explicitly not true of GNS: It is not about the way decisions are made, but rather what decisions are thought to be important, which certainly is prone to influencing the strength and priorisation of the filters that are used.

It will do little harm to think about the filters you use and of those that other people you game with use, and the order in which they are likely to be applied. Something might even be learned by such actions.

1. #### Opusinsaniasaid,

This sounds like a bunch of theories in (American) sociology, where they try to describe action as based (solely) on choice. It has its roots in economics and the theories of rational choice. It has also received a lot of criticism, which I consider to be mostly well-aimed. First and foremost is the presupposition of rational choice and complete psychological transparency. A theory similar to yours was questioned based on the assumption that humans are at all capable of forming such “preference filters” or at least in any coherent way.

What this means then? That as a heuristic model of play, yours will probably do some good. But it does not portray actual decision making very accurately.

2. #### Tommisaid,

A simple model of a complex phenomenon is always flawed (by the definition of a “complex”). My model does not assume that filtering is always done consciously, though, which negates some of the conserns.

Maybe I should write an entry on the nature of models.

3. #### Josh Wsaid,

How about the filter maps elements of the decision space many to one to a partially ordered set, where an option is better than another in terms of that criteria. Then somehow you add those orderings together by combining the filters until you end up with a fully ordered set? In this one you don’t create full numerical values, just that a is greater than b, and then you say that if a>b in filter c, and ab. You could even add degrees of “betterness” expanding it, and bringing it closer to a real number line. I suspect this more flexy model is closer to the way people decide. YMMV

What decisions are thought to be important is in itself a decision isn’t it? Can you treat it the same way at a different level?

4. #### Tommisaid,

Hello Josh.

Your idea sounds intriguing. At the time of writing this I knew little about (partially) order sets; my knowledge is limited still, at least to the level that I can’t really call your approach intuitive.

What decisions are thought to be important is in itself a decision isn’t it? Can you treat it the same way at a different level?

I don’t see a reason for not doing so. My general approach is to start waving my hands and saying that play preferences are given. At least currently I am not very interested in researching them further.

5. #### Josh Wsaid,

Fair enough, have you ever tried to order food and known that you like this more than that, but you can’t decide which of these other two is better? So it’s not that they are equal, it’s just that their relative goodness is undefined. That’s what I’m going for. You start with something more nebulous than the number line, and if your not careful, you can produce loops like a is nicer than b, b is nicer than c, c is nicer than a! But on the positive side you only ask people to do comparisons they think they can do.

So what if you find someone does have a preference loop? It seems pretty unstable! I’ve seen this happen when people’s filters are pretty incompatible, but they only ever apply them separately. So perhaps the way to do it is to add an extra distinction, “this is better than that when ___” and add detail to the model.

6. #### Tommisaid,

Partial order: IIRC, partial order is transitive and was it antireflexive (aRb and bRa imply a = b). Your example seems to attack transitivity. Again, IIRC, partial orders are not defined between all pairs of relevant objects; for example, “subset of” is a partial relation. So take $X = \{1,2,3\}$. Now the relation is not defined between, say, $\{1,2\}$ and $\{2,3\}$.

Loops: The order in which people apply their filters is, I think, one of the more interesting questions.