I’d argue it’s the opposite reason: rigid-body physics can theoretically be implemented by a high school student (getting collisions right would be an agonizingly long process, but all the math is within their grasp). But consistently passing the Turing test, much less with a full range of human emotions and the ability to react to novel situations, is still an active area of research.
In other words, a normal conversation is significantly more complex than even top-of-the-line physics engines! So it’s easier to have actual humans write scripted responses than to make a general system for it.
I don’t think I set the bar this high. Of course full AI is difficult. I think I was responding to the idea that NPCs react in a systematic or mechanistic way, above and beyond conversation menu choices.
For the most part, the technology used in the mainstream hasn’t advanced much past the CYOA stage as found in games like Baldur’s Gate 25 years ago. Pick the right pre-scripted option, win the girl. This is as much a sign of the limitations of the technology as it is a sign of the buying power of the market and the attitudes of (mostly white male) designers.
But the hobbyist IF market is much more diverse, and still tends to stick with either topic lists or conversation menus. Even the ones that use emotional state systems (Galatea and Alabaster are the two that come to mind first and foremost) tend to follow one of those two models, with pre-arranged lists of options (keywords or choices) and pre-written scripted responses to those options.
I’m pretty sure it’s not because the IF scene is dominated by white cishet men, it’s because this is the model that’s proven most effective with the technology we have. There have been some really cool advances in this field in the last few years (like Spirit AI’s Character Engine), but the best pre-written responses are still miles and miles ahead of the best automatically-generated ones. Compare your average conversation with Alexa or Siri (modern commercial attempts at creating natural responses on the fly) against your average playthrough of Galatea (decades-old freeware with scripted, pre-written responses) and the latter will win out every time.
Do we even really want AI in our fiction games? It’d be great for maybe an MMO or something where you just want a believable world, but I think it would be pretty bad in a storytelling environment. All I can picture is every time a DM’s carefully planned story completely fell to pieces because the D&D players just did whatever they felt like. Even if they play by the rules, it’s nearly impossible to tell a well crafted story in that environment. Or at least not one on par with an environment in which the author has full control over every NPC’s responses (in this example, the other D&D players are the AI NPCs). The downside is that the responses will be finite and predetermined, but the tradeoff is worth it, IMO.
Non-sequitur. I’m not talking about conversational AI or computer-generated responses. That’s beyond the scope of what we can do. The system I propose will still use human-written responses. It’s simply a way of keeping track of complex multi-dimensional emotional states.
This is a fascinating idea, i’m looking forward to seeing what you can do with it. Out of curiousity, have you considered modelling character predispositions as part of this system?
By that I mean, given the same set of initial emotional axis’, a persons own personality and psychological predispositions are likely to have an effect on attempts to manipulate them… for instance, a brutish thug conan has met in a thief encampment in Zamorra is likely to be easy to rouse to higher levels of agression, but harder to frighten… particularly if they’ve been drinking. On the other hand, a poet in the court of Aquilonia (save the unfortunate Rinaldo perhaps) is likely far easier to intimidate.
You could do this with some sort of weighting for each of the axis. There could be some interesting side-effects there too - if weighting was applied seperately to each end of the axis, you could have a character who is easily led for instance, by weighting both ends of “brave/timid” highly the character would become easier to inspire to bravery, or cow via intimidation.
Tone note: Conan doesn’t seem the type to “soothe” anybody. Seems like, in order to tell someone aggressive to back down without it coming to blows, he would instead tell a ribald joke about women to diffuse the anger, or talk of his and his listeners common foes to redirect the anger, assuming his listener is another meathead like himself.