Print

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

[Gadersd]

Has Tarn ever expressed any interest in adding evolutionary capabilities to the game? I think having animal features naturally adapt to each environment would add a lot of interesting variety to the game. Maybe even aspects of personality and mind can be tied to genetics in the game? I don't really mean this as a suggestion to the game. I'm just curious what Tarn's thoughts on the matter are.

[PatrikLundell]

Evolution generally happens over larger time scales than the DF history, and the biomes are largely fixed (although that will change to some extent with the map rewrite and Myth & Magic). However, it sounds like something he might be interested in, although that would then clash somewhat with the raws.

If you want Toady's thoughts on the matter you should post your question in the place meant for that purpose: the Future of the Fortress thread (pinned as the first thread on this sub forum). Note that questions are marked with (lime) green to allow Toady to find them when answering the questions at the beginning of each month.

[Sajiky]

I can think of maybe 5 (I'm sure a biologist could come up with many more) extremely minor evolutionary changes in organisms since the start of recorded history.

It is a really cool idea, but I think he'd have to code world generation on the level of epochs rather than years.

I don't know about you, but I don't have enough decades left in this life to run that world generation simulation.  As is, this game is clearly the deepest (and I think greatest, despite my failing every few years to get fully invested) game ever created.

[Gadersd]

I think the time scale may not be as big of an issue as you might think. Here is an article about rapid evolution of fish: https://phys.org/news/2009-06-evolution-years.html
I've done a few evolution simulations of my own and I've seen significant changes even in the course of a day. My simulations are quite different than what Tarn would likely implement, but I think they give support that this could be viable.

I can think of a few situations that might yield very fast evolution. Maybe an isolated area that becomes flooded up to 5 and a half feet. Most of the shorter people might drown or find life much more difficult than taller people who could walk through the water. This situation seems a bit contrived, but any significant environmental change that kills or hampers some of the population can lead to very fast changes in the gene pool.

People born with powerful magical ability would have much better survival rates than mundane people. I think this could lead to populations with very strong magical talents or maybe even magic resistance to survive.

All this may very well be out of the scope Tarn has in mind for Dwarf Fortress, but I think it would lead to interesting cases.

[Red Diamond]

Has Tarn ever expressed any interest in adding evolutionary capabilities to the game? I think having animal features naturally adapt to each environment would add a lot of interesting variety to the game. Maybe even aspects of personality and mind can be tied to genetics in the game? I don't really mean this as a suggestion to the game. I'm just curious what Tarn's thoughts on the matter are.

How about you head over to Future of the Fortress and ask him?

[Gadersd]

I just asked about it in Future of the Fortress. I'll have my fingers crossed. If Tarn isn't interested I might make my own game to try it out, although I'll probably do that anyway. I've never seen a game do evolution justice, but we might have to wait for quantum computers to come up to speed.

[Red Diamond]

I just asked about it in Future of the Fortress. I'll have my fingers crossed. If Tarn isn't interested I might make my own game to try it out, although I'll probably do that anyway. I've never seen a game do evolution justice, but we might have to wait for quantum computers to come up to speed.

No game does it justice because it does not actually work and few want to face the non-workingness of evolution.  To take your earlier scenario.

I can think of a few situations that might yield very fast evolution. Maybe an isolated area that becomes flooded up to 5 and a half feet. Most of the shorter people might drown or find life much more difficult than taller people who could walk through the water. This situation seems a bit contrived, but any significant environmental change that kills or hampers some of the population can lead to very fast changes in the gene pool.

People born with powerful magical ability would have much better survival rates than mundane people. I think this could lead to populations with very strong magical talents or maybe even magic resistance to survive.

All this may very well be out of the scope Tarn has in mind for Dwarf Fortress, but I think it would lead to interesting cases.

This situation works about one time.  You drowned everyone below 5/2 feet and now everybody is taller than that height.  Now you introduce a deadly gas cloud that hovers 5 feet above the ground and no evolution is going to happen at this point; everybody just died and that was the end.  The only reason that evolution worked in that particular instance is that the absence of selection prior to that point allowed a diversity of heights to exist (it did not matter how tall you were).  Now after you selected for height, everybody is tall so there is no ability to adapt to your new selection for shortness. 

Evolution requires both diversity and selection; selection kills off diversity. 

[Gadersd]

I'll just have to make a proper game to demonstrate. I've done simulations before so I'm personally convinced that it'll work at least to some playable degree.

[Starver]

I'm sure diversity (mutation, asexually; allele frequency shifting, for polyploid versions of the Raws) can be arranged. But I'm less sure that a sane selection criteria can be crafted.  Without 'preloading' some large assumptions.  That's what kills a lot of so-called-Evolution simulations.

Actually, the one I think I liked best did the latter by crowdsourcing.  Darwinian Poetry had a simple 'breeding' system (it cross-cut poems consisting of elements A1A2A3A4A5A6... and B1B2B3B4B5B6... at random intersections to create two 'children' at a time, like A1A2B3A4B5B6... and B1B2A3B4A5A6...) and left the issue of selection (comparing a random pair and upvoting one, downvoting the other, and peridiocally culling the downvoted ones before breeding a new generation to add to the survivors) down to the whims of the visitors, who could be individially capricious/random but (as a 'crowd in the cloud') produced some good developments. Or allowed them to be produced, rather.



Now, I think that with a lot of work we could get creature-Raws into a format that can allow an allele-type analogue (for similar-enough parents, though we could fulfill the half-breeds/etc of another recent request), but I fear that the additional burden of sensibly allowing diversity to thrive and then allow a sensible (game-relevent) selectivity would create a worldgen/continuation-of-history process that was far more resource intensive and just as likely to not turn out how you want it to anyway...


Right now, I don't think DF is the game for this.  Randomised tropes in the setting up of various (currently) fixed elements, maybe, as is apparently highly featured in future aims, but that isn't true evolution, and true evolution (or even a half-assed mockery of it) has the potential to be more dissapointing than the proposed permutation of building blocks only at creation, and seeing where that sends History.

[Red Diamond]

I'll just have to make a proper game to demonstrate. I've done simulations before so I'm personally convinced that it'll work at least to some playable degree.

There is a difference between working in general and working in a way that makes for good gameplay.  What gamers actually want is progress, which the ignorant think is the same thing as evolution.  They want things to get better, faster, stronger, cleverer etc over time because that makes them happy.  What evolution actually ends up producing is things like whales, creatures that have lungs rather than gills despite living in the water and dying if they try to leave it. 

So the fish evolved lungs to live on the land and lost their gills.  Then said land-animals evolved back into something that looks like a fish but has no gills; so overall they ended up worse off than they were to begin with.  Gamers do not want a game where the outcome of successful gameplay is they are worse off than if they never played the game in the first place, but that is a common outcome of the evolution-game.

I'm sure diversity (mutation, asexually; allele frequency shifting, for polyploid versions of the Raws) can be arranged. But I'm less sure that a sane selection criteria can be crafted.  Without 'preloading' some large assumptions.  That's what kills a lot of so-called-Evolution simulations.

There is no need to predefine selection, selection just exists as soon as you add a non-random 'lose condition'.  In present DF the 'fittest' in terms of natural selection are the necromancers and in terms of artificial selection the dwarves (because they are playable).  You could say the player is artificial selection working counter to natural selection; which ultimately means the necromancers.   

Selection is also largely a bad thing for evolution.  Evolving something by selection is rather like taking a book, ripping out half of the pages and then declaring that you have a new, better book.  You run out of pages and then you are stuck with a uniform creature perfectly adapted for an environment that just changed into something completely different. 

Darwinian Poetry is deeply flawed as a model because the selection happens a lot faster than the diversification does and the latter is localised unlike the former.  Keeping the poem the same is an option in real-life but the game encourages constant change to the poem, unlike real-life.

[Starver]

(Not sure I agree with your assesments on the Poetry. No poem is changed, it merely mates and produces (oppositely reconfigured) twin babies at each match. It survives as long as it takes until selection deems it sufficiently worse than all other, and I remember that 0^th-Generation poems lasted several cycles of cull'n'(pro)create, and us afficionados of the site enjoyed seeing how long an n^th generation contender actually lasted. And with no senescence factor (and the possibility of truly dire descendents from half-decent ancestors) some lasted a surprisingly long time. Though "oh no, it's that one again", or eventual dislike of an oft-reproduced gem of a phrase, ight have produced downscoring. Anyway, there's many things different from DF, forgive my nostalgia dragging it in as an imperfect exemplar.)

[Gadersd]

There is a difference between working in general and working in a way that makes for good gameplay.  What gamers actually want is progress, which the ignorant think is the same thing as evolution.  They want things to get better, faster, stronger, cleverer etc over time because that makes them happy.  What evolution actually ends up producing is things like whales, creatures that have lungs rather than gills despite living in the water and dying if they try to leave it. 

So the fish evolved lungs to live on the land and lost their gills.  Then said land-animals evolved back into something that looks like a fish but has no gills; so overall they ended up worse off than they were to begin with.  Gamers do not want a game where the outcome of successful gameplay is they are worse off than if they never played the game in the first place, but that is a common outcome of the evolution-game.

The evolution for the game idea in my head has little to do with absolute progress. I'm personally more interested in the interesting alterations the creatures would go through when exposed to wildly varying environments and conditions. I would be ecstatic to discover a land animal slowly changing into an aquatic form or vice versa.

My goal is to have the aspects of the creatures' bodies and behavior change and adapt in a natural feeling way. The mutation rates can be fiddled with to speed up or slow down the rate of diversification if needed.

Such a game, if it would even be game at all, might not be fun for most people. It's just an idea I have a personal investment in. Even if I must keep 95% of traits hardcoded and fixed to get good results, I'll be satisfied.

I think the idea has most merit for games with a few parameters that the developers are unsure of. Maybe in a strategy game different groups can evolve certain traits such as an inclination to build defenses rather than focus on attack power alone or preferring certain weapon types over another. This has the possibility to improve or diversify the AI of games. I have seen projects that use evolutionary algorithms to auto balance strategy games. There is a lot of untapped potential here.

[Nopenope]

Evolution by random drift is pretty much already in the game. Natural selection kind of occurs when the murderous player hunts populations to death, but there's no telling whether that actually fixates the 'alleles' in the (very crude) in-game genetics system.

[Red Diamond]

The evolution for the game idea in my head has little to do with absolute progress. I'm personally more interested in the interesting alterations the creatures would go through when exposed to wildly varying environments and conditions. I would be ecstatic to discover a land animal slowly changing into an aquatic form or vice versa.

My goal is to have the aspects of the creatures' bodies and behavior change and adapt in a natural feeling way. The mutation rates can be fiddled with to speed up or slow down the rate of diversification if needed.

Such a game, if it would even be game at all, might not be fun for most people. It's just an idea I have a personal investment in. Even if I must keep 95% of traits hardcoded and fixed to get good results, I'll be satisfied.

I think the idea has most merit for games with a few parameters that the developers are unsure of. Maybe in a strategy game different groups can evolve certain traits such as an inclination to build defenses rather than focus on attack power alone or preferring certain weapon types over another. This has the possibility to improve or diversify the AI of games. I have seen projects that use evolutionary algorithms to auto balance strategy games. There is a lot of untapped potential here.

A big problem with for instance whale evolution is the lack of intermediaries because creatures abilities tend to be binary; a creature is aquatic or a creature is terrestrial because of it's tags.  If you simply take the land animal and throw in in the water or the other way around the creature simply dies.  Biomes are also in the same boat, something either is a desert or it isn't, semi-desert for instance is not a thing as it is in real-life.  The absolute binary nature of a computer game's optimal coding will thwart evolution rather well I suspect.

The player need for "progress" is not to be dismissed however.  Evolution of the AI runs into the possibility the AI might end being too clever, the player does not feel they are learning to play the game at all because the AI has simply learned how to beat the player faster than the other way around. 

Evolution by random drift is pretty much already in the game. Natural selection kind of occurs when the murderous player hunts populations to death, but there's no telling whether that actually fixates the 'alleles' in the (very crude) in-game genetics system.

It doesn't, which is why the player can so easily hunt populations to death.  If it did happen the creatures would sometimes become too cautious and stealthy for the player to find rather than dying out.  The game makes minimal use of genetics for things like skin colour and hair colour, basically cosmetic stuff.  Personality is decidedly not genetic, it is random as is the abilities of the creatures.  It seems almost like the game design been carefully avoiding avoiding eugenics and animal breeding. 

The one instance of actual evolution in the game is actually the rise of the necromancers.  Necromancers are an evolutionary descendant of their parent creature in a fashion and they are also fitter than their parent creature, such as the ultimate fate of most dwarf fortress worlds is to be take over by them. 

[Nopenope]

  If it did happen the creatures would sometimes become too cautious and stealthy for the player to find rather than dying out.

That's not really how evolution works. What you're describing is differences in behaviour which are largely set by fixed tags like [LARGE_PREDATOR], [BENIGN] and so on. And yes, species that do not fear the player may eventually go extinct just like they did in real life (see the dodo bird or one of many predator-free insular animals really). So, in a way, the presence of a murderous player naturally selects for fearful and fast creatures. Selection just happens at the species level rather than the intra-population level.

At least, in theory. In practice, just like in real life, evolution is largely dominated by random drift. The way megabeasts gradually go exctinct despite having much better fitness (stats) than most other species is almost a textbook example of drift dwarfing selection in magnitude.