Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Makbeth]

I, and most likely nearly all of you, constantly see the Dwarven Index rear its head in my fort.  For those unfamiliar with it, the index is a segment of a D for Dwarf article on the wiki that states that a dwarf making the worst of two or more decisions is more probable than random chance would suggest. 

Now, D for D articles are meant to be taken with a grain of salt, but I've started wondering if this was intentional on Toady's part, in order to increase the difficulty.  These guys have a supernatural ability to narrowly avert success at the last moment and endanger themselves and the fort, or at least hold it back. 

Do you suppose this is intentional?  I always thought it was because the game is in alpha and the AI was only made functional enough to be tolerable for now.  But if that were the case, I would have expected an error rate closer to 50% and right now the dwarves definitely seem to favor failing strategies over winning ones.

[Max White]

Dear god I hope so!
I mean there is a good chance that this is simply because we don't notice as much when a dwarf measures up to standard, we expect it, while when a dwarf screws up, we take note. So when you tally times when a dwarf does well to when he does poorly it looks like the dwarf is dumber then he really is. However, if the AI is programmed to make the wrong choice, that might be just enough to bring this game up to the next level of win! And taking into account that DF is already a good 55 levels higher then the max, that takes a lot.

[Greiger]

My own personal theory about it is that theres a bit of a bias on random choises.  Just like how dwarfs prefer to build a wall from the west instead of the east or north over the south.

For example I'm thinking when scared by an opponent they flee away from their enemy.  We know this happens. And it almost happens intelligently as long as the dwarf has a path directly away from the enemy.  They will flee more or less away.  Regardless of whether or not that direction is actually the smartest path.  They just want to get far away, they aren't doing any complex pathing.  Its just, enemy to the east?  Flee west.  Which is more or less understandable in most situations, it's panic.  And opposite of the threat is the most obvious answer.

The problem comes when you put a wall there.  Suddenly they don't have a direct path away from their enemy, then the dwarven index kicks in.  The RNG picks a direction to go.  However unlike what many others think, I don't believe the direction they flee in that case is entirely random.  I'm thinking just like walls, they prefer to flee to the north or west.  Providing the threat that triggers the panic or something that blocks pathing is not in that direction.  Which could explain not having a 50-50 division.

Under that thought a north facing or west facing 'path to safety' could actually result in a higher apparent stupidity. than a south or east facing path.


Oh and also likely contributing to it, they seem to completely ignore stairs(if not z-levels altogether) when panicked.  I've had dwarves standing on stairs to safety and still flee straight into the hordes because apperantly they thought they were in a dead end when running the panic pathing.

DISCLAIMER: All of the above is all based on casual observation combined with speculation.  I'm too lazy to apply !!SCIENCE!!.

[Granite26]

also, a dwarf encountering an enemy that will chase it is likely to be returning to the fort.  That means that most of the time, a dwarf encountering an enemy will have that enemy closer to 'safe' than not.

[LShadow]

I think that the possible number of "bad" decisions outnumber the possible "good" decisions. If one is randomly selected, it's just more likely to be a "bad" one.

[Anathema]

I think that the possible number of "bad" decisions outnumber the possible "good" decisions. If one is randomly selected, it's just more likely to be a "bad" one.

I'm going with this one. To use the "dwarf running from an enemy" situation as a basic example - there's probably one, and exactly one, "path to safety" that will get the dwarf to your trap-laden entrance and shred the enemy (after all, if you went around scattering dozens of entrances to your fort across the map, it'd be hard to keep the fort secure. Unless you then routed these scattered secondary entrances all to a well defended underground main entrance.. hmm.). But anyway, there's probably one path to safety, and on the other hand dozens of ways to run that do not lead to safety, resulting in a messy death. Dwarves aren't deliberately being stupid (i.e. avoiding the "path to safety"), they're just not smart enough to recognize that one correct choice, so they make a somewhat random choice based on the happenstance positioning of the enemy and other obstacles, and the availability of many poor choices vs. one good one means it's a rare dwarf who stumbles across the one right choice by luck.

It applies to many other decisions if you think about it - there's only one good choice you want your mechanic to make, hook up the lever to the drawbridge before the goblins get here, and plenty of less-than-optimal things he could do instead (eat, drink, sleep, make more mechanisms, collect sock..), so the odds are just against getting what you want. Or you want the hauler to get that miasma-spewing corpse out of your trade depot, but there are hundreds of other things he can haul instead. It's possible for the experienced player to tip the odds a little in your favor by eliminating bad choices - forbidding the sock and suspending the mechanic's workshop, or setting at least one hauler to only do refuse/burial - but in the end it's a complicated game full of less-than-optimal choices for your dwarves to make, and you just can't prevent all these bad choices.

I think if you pay close attention to the very few situations where there's only one bad choice to balance out the good one - like the classic example of walling off an area, usually there are only two sides of the wall to stand on - then you'll find it generally does come out to 50/50, the bad decisions are just more noticeable/memorable. In fact, for very simple decisions like this, it's possible to gain some insight into how the dwarf makes the decision (i.e. mason prefers to stand north of the wall) and even bias things to have a more-likely-than-not shot at getting the good choice. But simple decisions like this are the exception - DF's complexity ensures that most dwarven decisions are far beyond any easy understanding of how they're made, and bad (or, at least, less-than-optimal) choices far outweigh the good one(s).

..That being said, it's a hell of a lot more entertaining to let yourself believe that they're being deliberately contrary to your fortress's success and their own well being. It's not very dwarfy, after all, to do things the easy way..

[LionSilverWolf]

I would have to say that it is a combination of Anathema's and Greiger's theories. Dorfs do, in my experience always, prefer to do things (anything) from the north/west. That, combined with the sheer number of ways to screw things up, skews the odds drastically in favour of things getting screwed. Now I'm off to incite a horde of gobbos to my fort so I may do !!SCIENCE!! to this theory.

[ZeroSumHappiness]

Also, you have to remember that if the Dorfs do the right thing 99 times you don't notice because your fort keeps on keeping on, but that one time they screw up it's tit's up and you notice.

[Aramco]

I think that the possible number of "bad" decisions outnumber the possible "good" decisions. If one is randomly selected, it's just more likely to be a "bad" one.

Options--
1. Step into magma.
2. Place hand in magma.
3. Stand still and get engulfed by magma.
4. Swim in magma.
5. Casually ignore magma until engulfed.
6. Taste magma.
7. Attempt to fight magma.
8. Start up a conversation with magma.
9. [Insert countless more stupid magma-related decisions here]
10. Place Elf in magma, then leave.

As you can see, the number of incorrect decisions vastly outweigh the one logical decision. Point proven correct.

[Cruxador]

Also, you have to remember that if the Dorfs do the right thing 99 times you don't notice because your fort keeps on keeping on, but that one time they screw up it's tit's up and you notice.

I don't usually notice either, because I arrange things such that dwarves can't really screw up in a way that I care about. They are allowed only to endanger irrelevant things, like their own lives.

[DStecks]

-snip-
As you can see, the number of incorrect decisions vastly outweigh the one logical decision. Point proven correct.

My problem with your theory, and Anathema's, (I quoted you because you typed less) is that you're assuming that the dorfs' decision making process is largely, if not completely random. The number of possible wrong decisions doesn't matter one bit if decisions are made based on anything but random chance.

[Untelligent]

I think one of the reasons it seems like dwarves tend to make stupid decisions a lot is because we never notice when a dwarf makes a "good" decision, just a "bad" one.

[Aramco]

I think one of the reasons it seems like dwarves tend to make stupid decisions a lot is because we never notice when a dwarf makes a "good" decision, just a "bad" one.

But how can a Dwarf make a logical decision? You try drinking nothing but booze and doing something right.

[Anathema]

-snip-
As you can see, the number of incorrect decisions vastly outweigh the one logical decision. Point proven correct.

My problem with your theory, and Anathema's, (I quoted you because you typed less) is that you're assuming that the dorfs' decision making process is largely, if not completely random. The number of possible wrong decisions doesn't matter one bit if decisions are made based on anything but random chance.

Well, I don't necessarily mean to imply that the decision process is "random," as in the game produces a random number and goes from there, it's just that it seems random. Any sufficiently complicated system you don't understand can seem random - I mean, there are no truly random events in our universe, you could theoretically calculate the result of a dice throw if you knew every possible variable about weight, distance, force, air resistance, elasticity, etc., it's just that the calculations are so complicated and so prone to change drastically from even small variations/inaccuracies in the starting conditions that it's not practical to predict it - so it seems random. The word "random," as we commonly use it, really just refers to things that are "too difficult for us to predict," which many dwarf decisions qualify as. By the way, read up on chaos theory if interested, it goes into more detail about how many simple predictable factors can add up to seemingly random (i.e. unpredictable) results.

And the number of wrong decisions does matter very much even if the decision process is not random. Say, for example, that when choosing items to haul the dwarf always picks the closest one. If you want him to haul a particular corpse, and that corpse is the only thing that can be hauled, you're good - if the corpse is one of 1,000 items slated for hauling, odds are it won't be the closest item. See, even though the Dwarf is applying a logical and practical method to choosing the item to haul, there's an element of randomness, i.e. where the item happens to be in relation to the dwarf. The astute reader will note that this is not really "random" either, but like the dice example it depends on so many variables - all the items that need to be hauled and how they got wherever they are, whatever job your hauler did previously and where it left him, all the possible paths connecting said dwarf to said items you may or may not have dug/constructed, and so on - so like the overall dwarf decision process, it would appear somewhat random to us who don't know every detail of how it works or every variable that affects it.

[ZeroSumHappiness]

At this time we believe that certain quantum effects and radioactive decay are truly random.