Return to “Suggestions”

Post

Re: NPC Characteristics

#16
else we get AI Darwinism, where the "fittest" (or probably, the most aggressive) AI eats all of the lesser ones over the course of a few days. I imagine the game to become incredibly unfun when all you encounter is xenophobic megafactions that have adopted a "kill on sight" philosophy on anything that doesn't belong to the faction, and is large and willing enough to throw hundreds of ships at the player.
This in particular is the causative factor that killed off a lot of "realistic faction-based" approaches that were tried out for games in the past (ranging all the way from Freelancer to Oblivion). The combination of a) certain strategies (expansionism, xenophobism) being unilaterally superior from a metagaming perspective, combined with b) the inability of other factions to adopt proactive, as opposed to reactive, approaches to diplomacy means that a lot of "strong AI"-like systems end up deliberately handicapped by randomness, or (more commonly) removed altogether for playability reasons. Don't want to see the same fate here.

For this particular scenario, one way to preserve at least some balance would be some measure of threat -- even though a xenophobic megafaction would certainly win against a minor fringe faction, the lack of threat combined with a (perceived) lack of reward, based on their intelligence*, makes them less likely to engage. That other megafaction in the corner of the galaxy, however... The same reasoning would apply to a player, of course.

Threat includes:
- Contested systems (most important)
- Proximity
- Total/rate of change of population
- Total/rate of change of military assets
- Position of military assets
- Cultural influence (ala SoaSE style)

Reward includes:
- Estimated value of current holdings
- Systems
- Populations
- Trade volume
- Technology
- Resources

*An interesting survival strategy therefore would be to present oneself as unappetizing of a target as possible. Intentionally nuke planets to create a desolate, radioactive wasteland, and instead have your uber gold mine deep in space somewhere beyond the fringe.
Post

Re: NPC Characteristics

#17
The traits that are most useful to people or psychologists may not be easily translated into code, particularly where these traits interact with a model that includes goods, weapons, resources, credits, locations, knowledge, factions, time and anything else that I missed.

Other traits that can be derived from (combinations of) the more fundamental traits may be more tractable to programming. A simple example is patience, a.k.a. delayed gratification. I imagine that psychologists would say that patience is an emergent trait derived from the combination of openness and conscientiousness. Economists often describe the ability to delay gratification as the most important economic attribute of people.

Patience seems to relatively easy to program into an AI with the characteristics that Josh has described so far. Actions have a duration and/or have a function of time to produce a result.

A patient miner will consider mining to fill storage before returning to a place to sell. Other factors such as (made up example) life-support and or energy resources may force even a patient miner to cut this short. On the other hand, an impatient miner might only mine enough to get a certain amount of credits before returning to base.
Post

Re: NPC Characteristics

#18
While a game universe is generated from a seed at the beginning of the game is there really a need for (NPC) activity outside of the areas the player has actually come in contact with? That is to say do trees need to be falling (much less heard) in a forest that the player has absolutely no information about? Couldn't it stay in its initial, pregenerated, state until the player either learns or makes his way to it?

While it seems that in Limit Theory the AI will be a cohesive unified whole I always assumed that it will apply different matrices and algorithms on the macro (system) level as opposed to the micro (interpersonal) level. On the macro level the issue of scaling was already broached and I have every expectation that for the sake of playability there will be any number of limits (boundaries, if you will) on the various entities (political, economic, military, etc...) that the player will encounter and or create.

On the micro level I see little point in generating interpersonal NPCs until they are encountered. In turn, NPC depth could directly correlate to the amount of interaction with the player. Such individual NPCs can be generated at the beginning of the game though I would think that to save resources it would make more sense to tie their existence to the approach of the player into a given quadrant, sector, system or potentially even a planet, space station, or asteroid field. I have nothing against the use of a personality (behavioral) matrix in general. I am suggesting however that it be used in conjunction with a stochastic simulation. In essence the model could provide potential pathways and boundaries but the NPC's personality would be generated through interaction in a probabilistic manner. As far as constructing a compelling narrative from the interaction, to a large degree players will presumably be able to rely on their own imagination. No need for Josh to come up with cathartic tales of woe or all's well that ends well merriment.
I know not what life is, nor death.
Year in year out-all but a dream.
Both Heaven and Hell are left behind;
I stand in the moonlit dawn,
Free from clouds of attachment.
Post

Re: NPC Characteristics

#19
Hardenberg wrote:Wouldn't it make more sense to construct NPC personalities around things that need to be done to keep the universe running and interesting? At the end of the day, we're still looking at a game, with the AI being required to facilitate a breathing, living universe.
...
I wish Josh would elaborate a bit on the actual mechanics of production chains, factions, entity spawning and the economy in general, since these subjects are what the AI needs to deal with on a regular basis. Once we have established what the AI can/must do, we can wax poetic about implementing personality along those lines.
I don't disagree with this... but it seems to me that starting from a core "personality" system is not incompatible with AIs that fill space with the interactions necessary for a game like Limit Theory.

In a way, we already know what the AI is "supposed to do" in LT to support human gameplay -- Josh basically told us that in the Kickstarter:
  • You are a Merchant.
  • You are an Admiral.
  • You are an Explorer.
  • You are a Pirate.
  • You are a Wingman.
  • You are a Miner.
Those activities are already pretty darn close to several of the basic motivations I suggested. I think I could make a decent case that Merchants and Miners are extroverted and introverted examples of the Security-seeking, economics-oriented play style; Wingman/Admiral and Pirate are about imposing Power; Explorer is about... exploring ;) ; and although it's not listed on the Kickstarter as a style of play, the part about "planetary ownership" at least lightly implies that high-level diplomatic gameplay that the social/story/person-oriented gamer can enjoy might also be a viable "career" in LT.

So designing NPCs to have those kinds of high-level goals, then unleashing them on the galaxy, helps to insure that there will be plenty of NPCs doing exactly the kinds of things that the LT universe has been said to support. They'll be out there flying/shooting and fleet-directing and mining and exploring because those actions are verbs that they're interested in applying.

There's nothing that says LT couldn't spawn NPCs to fill specific role slots, with no personality needed. An explorer exists only to explore, and so on. That would certainly be simpler.

I definitely think there are gamers who'd be just fine with a universe of cannon fodder, as in many existing games. I'm not one of them. I'd like NPCs in LT to have some personality behind their actions (even if some of them are just going to be shot down), I think there's gameplay value in that, and I think it can be done at a relatively low cost in development time and run-time performance.

But of course I'm not the person doing the actual work. :) I look forward to seeing how NPC behaviors wind up being expressed.
Post

Re: NPC Characteristics

#20
Flatfingers wrote:In a way, we already know what the AI is "supposed to do" in LT to support human gameplay -- Josh basically told us that in the Kickstarter:
  • You are a Merchant.
  • You are an Admiral.
  • You are an Explorer.
  • You are a Pirate.
  • You are a Wingman.
  • You are a Miner.
A possible problem that could happen with this approach is that an NPC could run into a dead-end when local conditions change, such as a key resource running out. A deeper and more abstract set of motivations might allow the NPC to evolve their plans in the face of a disaster.
Post

Re: NPC Characteristics

#21
jimhsu wrote:If possible, I'd like to borrow from psychiatry and use one of the models out there (OCEAN (the Big Five), Myers-Briggs, etc). It would definitely fit more towards Josh's vision of increasing levels of abstraction.

Taking OCEAN for example:
See http://www.economist.com/news/science-a ... lanet-apes for an example of an alien personality dimension - Chimpanzees have seem to have a sixth personality dimension that is not present in humans.
Post

Re: NPC Characteristics

#22
Ultimately these are all valid approaches I think. Anything that provides AIs with varied and interesting behaviours will do the job. Josh's description of each AI simply maximising value, where value is defined by the personality of the AI seems to be a quality approach. Throw in a some sort of weighting based on previous experience (or "skill" at a task) and you've got a really exciting system.

The only question to my mind is if and how this data should be transmitted to the player? Freelancer told you if the NPC was an enemy, neutral or a friend as well as his/her level. Do we want more? If we've got this fancy set of personality attributes governing how the NPC acts, it seems a shame to hide them from the player. Why not have a "Personality" screen that illustrates this stuff and allows you to get a handle on who you've encountered?

When you hardly know the NPC then the data you get would be necessarily incomplete. However if you belong to a faction then you could get an automatic reference against what is known by other members of your faction. As your familiarity grows then so does your own assessment of the personality. This data could then be sold... or purchased.

Combining personality traits with known faction allegiances may allow you to determine a "likely disposition" or some other rating that would allow you to guess if s/he'll attack or not.
Post

Re: NPC Characteristics

#23
mcsven wrote:Anything that provides AIs with varied and interesting behaviours will do the job. Josh's description of each AI simply maximising value, where value is defined by the personality of the AI seems to be a quality approach. Throw in a some sort of weighting based on previous experience (or "skill" at a task) and you've got a really exciting system.
When the AI demonstrates the emergent behaviour of a character or faction that maximises value by buying and selling information about other characters and factions then you know the AI and economics are fundamentally good.

One worry that I have is how to dumb down AI characters so that they are not super-competent to the point of overwhelming starter players. And less than competent players ;)
Post

Re: NPC Characteristics

#24
My turn to dabble in the necromantic arts. :)

I'm still extremely interested in the depth of the AI that characters in Limit Theory will have. And I still like the ideas in this thread.

But for the moment, just to poke some heat back into the embers, I'm going to cheat a little and repost something I recently suggested over at OtherSide's forum (for Underworld Ascendant, System Shock 3, and their just-released VR game, Underworld Overlord) regarding what I'd like to see in character AI for a game.

This is the package of AI-related features I'm waiting for a game to deliver:
  • Character motivations: what are this character's fundamental needs that drive general behavioral preferences?
  • Character abilities: what is this character particularly good at doing?
  • Character functional roles: within a community, what does this character normally do?
  • Character socio-economic relationships: with what organizations is this character affiliated?
  • Character personal relationships: whom does this character like and dislike?
  • Character schedule: what are this character's regular habits?
  • Character perception: how good is this character at forming a generally correct understanding of immediate phenomena and/or long-term dynamics?
  • Character reasoning: how good is this character at formulating effective and practical plans for achieving tactical or strategic goals?
  • Game objects: is there a wide and diverse range of objects in the world whose different functions characters can perceive?
  • Game verbs: is there a wide and diverse range of actions that characters can take for acting on objects?
  • Narrative framework: is there an outline of a good story whose plot points are triggered by appropriate actions/events?
Getting any of these alone would be nice. But it's when they combine as a package, interacting with and deepening each other, that I believe real magic will emerge. Not "I, the Designer, have contrived a specific experience for you, and if you try to engage with my code widgets in any other way, you are Doing It Wrong," but a clockwork universe set in motion by a designer, with truly variable outcomes based on the player's choices but all looping around a defined story framework that delivers a rationally or emotionally meaningful story.

...

There is nothing inherently wrong with scripting; I'm not anti-scripting. What I am is pro-dynamism: I think there's room for more games that rely less on developer control and more on procedural dynamics, where consequences for player actions are more interesting because they're more varied, and they're more varied because they don't all have to be meticulously defined and constructed by a developer, but emerge naturally. That doesn't mean you can't script some things -- it just means you don't have to do as much of it everywhere, giving you more time to build and polish your setpiece sequences that advance the main plot of your central story.

I'll readily agree that this kind of game would be harder to design, harder to build, harder to test, and harder to sell than the developer-centric games that have been made for 30 years.

But I won't agree that no one should try it.

What do you folks think about these suggestions? Are there other aspects of character AI that could help shift some of the dynamic fun of game creation to the characters themselves, offloading some of that work from the developer?
Post

Re: NPC Characteristics

#25
This kind of reminds me of Rim World where the NPCs have a degree of free will. I want to see more games focusing on cause and effect, with fewer generic game elements. Most games lack depth and interesting systems which work together to make a game fun.

There were a number of games made in the 90s which are still amazing because no one has been able to obtain that level of depth.
Image

Online Now

Users browsing this forum: No registered users and 19 guests

cron