Nov 11

We think we know what’s happening, but really we don’t

In the beginner’s mind, there are many possibilities.  In the expert’s mind, there are few.  –Shunryu Suzuki

Although we may think we know what’s going on around us, our brain is guessing.  We don’t get much input from our senses to tell us, for example, what’s physically behind us, but we have a pretty good guess that if there was a Christmas tree behind me 5 minutes ago the Christmas tree is probably still there.

I’d like to call our brain guessing about the world our “mental models.”  I have a mental model of the room that I’m in, that helps me pretend in my brain that I know what’s going on in this room, what will go on, and what has gone on.  If I can be computer-y about it, it’s like I have a function


where I can guess, based on a state (e.g. night-time, no one’s at home) that I know what’s going on in the downstairs living room, or at least the boundaries around what’s going on in that room.  (In my mental model for this room there could be cats clawing our couches, or a robber, or nothing, but there’s probably not a clown dancing.)  In a similar way of guessing what’s going on, I may have a mental model for our cat Fred:


where I can guess what our cat Fred might be doing based on what’s going on (e.g. how much food is left in the bowl, whether there’s a Christmas tree).

A few years ago, f_fredthecat(Christmas Tree) would have yielded in my brain “Fred is probably climbing the Christmas tree.”  He really liked climbing our Christmas trees and could climb two or three feet off the ground.  Now, though, he’s older, and f_fredthecat(Christmas Tree) tends to yield “I may find pine needles in the litterbox.”  My brain’s mental model for Fred f_fredthecat changed as a reflection of a change in Fred.

Minds tend to like guessing about what’s going on in the world, and I’d bet we each have hundreds of thousands of mental models.  For example right now I have mental models like…

  • f_myhomelaptopcomputer: this laptop’s plugged in and charging, its brightness is near maximum, its hard drive has plenty of space, it’s been backed up recently by Time Machine
  • f_jeans: my jeans are dry, they are comfortable
  • f_kitchensink: the kitchen sink is probably dripping right now although I can’t see it or hear it
  • f_weddingpotteryplate: the plate made by a family friend at our wedding is still balanced on our plate rail (I can see it), and even though it’s been there for several years I am still convinced it will fall soon
  • f_highchair: the high chair’s wheels are locked in place (untested)
  • f_workbuilding: Six months ago I would have thought our work building was empty.  Three months ago I went in at this time of night and saw a cadre of custodians cleaning and vacuuming, so now I’ll guess there are probably lots of people cleaning up.

All of these things are guesses–possibly very educated guesses, but they are guesses.  Although my laptop charging light is on, it is theoretically possible that power is not actually going into my computer.  (Aside: Can you tell I don’t understand how charging a battery works?)  As another example of a possibly false guess, although my jeans feel dry they could be getting wet and I haven’t noticed yet.

There are a couple of challenging things for me about mental models:

  • They are not real.  I think I know all these things, and that I know what’s going on around me, due to a small number of observations and a LOT of historical experience, reading, similar things, etc..  All of these mental models are reduced/simplified versions of reality.  The kitchen sink may have stopped dripping–it’s unlikely but possible.  The high chair’s wheels may be crawling with ants.
  • These models are unique to each person.  You may have a well developed mental model for how cakes bake in an oven.  Someone else, say a professional baker, may have another mental model.  These two expert models are probably very, very close to one another but they likely have small differences based in the different experiences of each person.  Yet we talk about “baking a cake” as if we all shared exactly the same mental model.
  • The models improve with interactions.  The more I’ve worked with something, the more experience I have, and the better understanding/guessing I can do in my mental model for the thing.  However (see next bullet)…
  • Past performance is no indicator of future results.  Just because the high chair has never exploded, does not mean it definitely will not explode.  My mental model does not account for the possibility that the high chair might explode.

The importance of first impressions

Mental models get really interesting for me when it comes to “new” things.  When I encounter something new, I search through my existing mental models to find an approximate match for the new thing.  For example, when I see a new chair, my brain searches through my mental models to find…

  • f_giantofficechair
  • f_uncomfortablechair
  • f_paddedchair
  • f_chairsthatcatssiton
  • ...

In other words, all of the different types of ways that I think chairs might work.  Then, when I sit down, I narrow down the range and eventually create a new mental model for this new chair that’s a mix of all my past experiences plus what I start to perceive.  Certainly, this is all done subconsciously, and is part of why I get surprised if the chair falls apart when I sit on it.  If that happened to me all the time, it would be part of my mental model.  Instead, if a chair fell apart when I sat on it my consciousness would wake up–“what the heck just happened?!”–and also my mental models for all chairs would change to include the possibility that the chair might fall apart.

Take all that and apply it to people.  I think that first impressions when meeting people are so important because you are sizing up the new person against all your other mental models of people.  You’re asking yourself who this new person is similar to.  This new person is “30% my eighth-grade teacher, 5% my best friend, and 65% that guy that punched me in the face.”  After more experience interacting with the person, that model would be refined, and maybe those foundational models are questioned–“oh, this person isn’t like my best friend at all”–but that mental model I’ve applied to the new person is, in my mind, actually that person, and it takes a good deal of work to realize my mental model was off.

On Guessing Effectively

So although mental models have lots of faults (such as being a poor caricature of reality although people think of them as reality), mental models are certainly useful.  They are the way our brain makes sense of the world.  I would guess that our brain has a core “function,” something like:

process_observation(observation, mental_models)

where when we observe something with one of our senses, that observation is filtered through our mental models, but then might update our mental models.  For example,

process_observation("I physically feel uncomfortable", {f_bed,...})

I feel uncomfortable laying on this bed, which reinforces my mental model about our bed, that it has an old mattress and needs to be replaced.  Or,

process_observation("I heard a squeaky noise", {f_bed,...})

if I heard a squeaky noise, I might guess it’s a bed spring popping, even if something totally different squeaked.

Basically I’m trying to say that as we hear, see, smell, taste, touch, or think things, these observations go through the filter of our mental models, but then have the chance to update or change our mental models.  We could perceive things “incorrectly” if our starting mental models are bad, or we could make poor conclusions based on how good we are at making better guesses.  All this to say, you have a big advantage if you are pretty good at updating your mental models from processing observations.  And you have a huge disadvantage if your mental models are so bad that you aren’t able to see they’re wrong and invalidate them based on observations incorrectly filtered through those bad mental models.

The better you are at guessing how things work and inferring how they work from your limited observations, questioning your mental models when they don’t make sense, the better you’ll do at approximating reality.  You will be able to make more lasting inferences and “predict the future.”  Mental models are not reality, but if you have skill at developing accurate mental models quickly then you will be “ahead” of reality getting your observations confirmed rather than running into input that contradicts what you think.

Nov 11

Why people play “Kingdom of Loathing”

This post is about a game, Kingdom of Loathing (KoL), and what I’ve learned about how the game works and how it motivates people to keep playing.  Despite this dry analysis, the game itself is very funny.

I have a “systems approach” to games, which I think means that I play them to understand the game dynamics.  However, although I’ve played for a while I am not an expert–as you’ll see below, this game is designed so that you could spend months of solid effort trying to understand how the game works.  This game is a trap for systems thinkers.

KoL is a “free, comical RPG.”  You can be one of six character classes: two are muscle-types, two are mysticality-types, and two are moxie-types.  For example, one myst class is a “Sauceror” that can cast spells like “Stream of Sauce.”  You have three corresponding primary stats, Muscle, Myst, and Moxie.

There are two big signals that this game is complex:

  1. It has a wiki, TheKolWiki, that currently has 8,599 wiki page entries including a chart of the best food to eat by main stat gain.  (Side note, “old school” wikis such as the original WikiWiki used StudlyCaps to indicate wiki pages.  I am guessing it is not a coincidence that TheKolWiki is written in StudlyCaps.) 
  2. There exists a front-end client, KoLMafia.  KoLMafia is designed for advanced players.  It automates many parts of the game such as adventuring, crafting materials, and shopping.
The high-level game looks something like this:
  1. Create a character and choose your class
  2. Level the character and complete all quests until you complete the final quest and defeat The Naughty Sorceress
  3. Gather other resources if desired, e.g. farm for meat (the currency) or items
  4. “Ascend” (i.e. start over), gaining some sort of benefit for ascending.  Optionally, choose to make your next game more difficult to receive additional benefits.  You can choose a different class.  Go to step 2.


Here are the high-level rewards (as I understand) for playing–the things that motivate people to continue playing:

  • “In-club” rewards such as the “haiku” chat channel for people who complete a haiku challenge in the game.  By “in-club,” I mean you are recognized as being cool or part of an elite sub-group.  There are hundreds of these types of rewards–everything from access to special chat channels to membership in prestigious clans like “Noblesse Oblige” to special items like “phish sticks” awarded to players who had accounts during a phishing scam and then changed their passwords.
  • “In-club” knowledge, such as knowing it’s called “KoL” or that Hagnk’s Ancestral Mini-Storage was once hit by a chunk of comet.  This knowledge you can use on the KoL Forum, in KoL in-game chat rooms, or other places where KoL folk hang out.  This also relates to a fundamental element of KoL: puns and inside jokes such as the “Piano Cat” familiar or an in-game Soul Coughing reference.
  • Trophies and collections, to show people or yourself how much you’ve done in the game, or how much cool stuff you have.
  • Smarter ways to play for people who put the time and/or cash money into it.  As you ascend, you can “perm” skills that you’ve gotten with each ascension: skills like Mad Looting Skillz (+20% item drops).  You can also buy items with cash money, or with “meat,” that help you play.  Please see “equipment,” below.
  • Knowledge is greatly rewarded.  There are food and drink recipes you can learn, which give you more turns per day to play.  You can learn good secret places to “farm” for items or you can find some trick to level your character more quickly.  The game rewards knowledge, but also other players reward knowledge.  The game has self-selected players that like knowledge.  This is part of the “systems thinking” trap–this game is designed with facets within facets so that you would have great difficulty learning, for example, what’s the best item to make with a Tome of Clip Art.  In this example, you’d need to understand all the possible items you could create, their effects, why you’re using the effect and whether it would be cheaper to buy something instead from the mall (or if you just want something should you instead sell a created item to the mall), etc., etc.
  • Longevity.  As you’ve been around longer, you have the option of getting equipment that’s phased out.  There are also many things that you can only do a limited number of times per day, e.g. get filthy lucre from the Bounty Hunter Hunter, so only people who have played regularly for a LONG time have access to certain items.
KoL is designed so that additional elements are added all the time to bolster these rewards.  The Tome of Clip Art, which introduced dozens of new items and play possibilities, was added in fall 2011.   New trophies are created.  Smarter ways of playing (perhaps relying on new items or item combinations or new play areas) are introduced over time.


KoL has MANY, MANY options for items, grouped into categories including (but not limited to):

  • clothing (hats, pants, accessories, and shirts–if you have the Torso Awaregness skill)
  • weapons
  • familiars (creatures that help you fight, or farm for meat, or do a host of other things)
  • food and booze
How do you get this equipment?
  • monster drops: when you beat a monster, or when you use a skill on a monster such as Pickpocket.
  • “semi-rare” adventures: every ~100 adventures.  There are formulae depending on what kind of run you are on.  One of KoLMafia’s many features is that it tracks your next semi-rare adventure.
  • ultra-rare adventures
  • quest/adventure outputs: you get items when you complete certain quests
  • existing: you get gifts on your KoL birthdays or on certain other occasions
  • holidays: there are in-game holidays where you can get special items
  • gifts: you can give and receive gifts from people
  • Bounty Hunter Hunter items: you can earn “filthy lucre” (one per day) for completing Bounty Hunter Hunter quests, and when you have enough lucre (200 for the Manual of Transcendent Olfaction) you can get special Bounty Hunter Hunter items
  • buying it with cash money: there is a store called Mr. Store where you can buy Mr. A’s for $10 USD and then trade Mr. A’s for other items.  In probably the smartest monetization of gaming, most of the items you can trade for (sometimes called Mr. Store items) are replaced monthly.  For example, according to the wiki, in January 2005 (only) you could trade for an “orphan baby yeti.”  The only people who have these yetis in the game now are people who traded for them in January 2005, or whoever won first prize in a raffle in 2008, or whoever bought these items (with meat) from the people who had them.
  • buying it with meat: there is a mall, where players can set up stores and sell most of their items.  (Some items are marked non-tradeable.)  To get meat, you can buy it–$10 USD Mr. A’s currently sell for around 10 million meat–or you can earn it–150k meat/day is a good haul (for me).  (This means your meat farming earns you around 15 cents/day.)  Some people “play the market” with the KoL mall, by buying cheap items and then selling them for more money later.
Each type of equipment can give you bonuses, and there are outfits if you wear certain combinations of hat/pants/etc.  Mr. Store items help you much more than “free” items, with the exception of some ultra-rare items or special reward items.
Notably, when you ascend, you can choose to play in a “hardcore” mode, or not.  If you play in hardcore mode you don’t have access to any of your old equipment EXCEPT YOUR FAMILIARS.  I think this makes familiars extra popular, because “really awesome” players play in hardcore mode or some even more hardcore version of hardcore mode such as a Bad Moon run.


Even here, I can’t capture the high-level gameplay in under 1000 words.  The KoL designers are brilliant, and I am sure they make good money with how they’ve designed Mr. Store and the mall.  This game attracts really smart people, and keeps them hooked. I broke my addiction most recently by reading a great walkthrough that called for ~$500 USD worth of items for “optimal” play.  I have now broken my addiction twice (so far) but keep getting tempted to play again and learn more and write my own KoLMafia scripts and play the KoL mall and buy Mr. Store items that I can sell in a couple of years to others.

I guess my simplest conclusion would be, if you have some time to kill and like puns, try playing Kingdom of Loathing, but don’t fool yourself into thinking you can find an “optimal” strategy for playing the game.  It’s misleadingly complex.