aigan said:
Especially so in Last of Us part 2. Don't want to play sociopath mass murderess. Ellie is pure evil.

Did not play any TLoU, but the curse of realistic characters - once they try to represent real world persons, killing anybody just no longer works.
Contrary, for a game like Doom, there is no problem with that at all.

aigan said:
Just making sure they are really dead kan take a lot of work. Then getting rid of the body. Then Trying to duck questions as to not get hunted by relatives or police. Then everything that comes with all the new lies.

Yeah, we could make this a mechanic. But replicating real world stress is not necessarily fun. Similar to how collecting resources to craft a shelter for survival often isn't fun either. Or many quests like hunt 10 deer often feel more work than fun.

Reality is a terrible template in general, imo.
We better invent an imaginary, alternate reality. Similar enough to ours it feels meaningful, but different enough to level out ethical / moralic constraints hindering us to do what we'd like to do.
So i do not think video games are here to learn something about life, relationships, political correctness, ethics - etc. No. They are here to distract from all those problems. To have a good time, a sensation, an experience. The purpose is similar to that of drugs, loud music, or an arena of gladiators.
With that in mind i don't have a problem with killing 1000 NPCs for a pot of ancient gold. But i do notice when it starts to look ridiculous, which does effect the fun.

What we can learn from reality is things like laws of nature. We want to conserve energy, so players can't craft infinite resources out of finite input. We want the same rules for physics, so our imaginary world feels like real. We want to replicate natural processes like erosion so our world shows some depth, etc.
But that's just what i think, and my interests do not represent a majority very well.

Regarding companion relationships, that's something idk. I never manage to play those RPGs long enough, and i never encountered their built in dating simulators yet.
But it's surely possible to combine this with a forceful and primate depiction of society without looking silly.
E.g., you have mentioned Campbells hero and character archetypes. This would give a story like Star Wars maybe - quite low brow and shallow, but accessible to a majority and potentially successful. Star Wars or Indiana Jones do work quite well with some love story added to the mix of force and violence. It's just that neither topic is handled in some deep or realistic sense.
And it can work with some deeper, more realistic, intellectual story as well. Like Blade Runner maybe.

The video games industry just lacks experience and competence on those things, i think. But they're constantly improving. Their art becomes better and better in general, including stories.

The violence is no big problem imo. We just need to give it a meaning, when it becomes necessary.

aigan said:
I provided a big list of systemic games. How can you still ask what gameplay the systems provides?

My doubts are no meant as critique, with in intent to question your belief in systemic games.
I believe it myself, but i just complement all my beliefs with doubts, all the time.
I can't be fully convinced about some idea, because then it becomes just wishful thinking to me. I need the doubt to realize its limitations, to keep an eye on alternatives, to keep expectations low and realistic. That's just how i am.

But your list does indeed not convince me about a systemic idea, since i know only a few games.
Also, and that's important - ‘systemic’ as i got it is much more a philosophy than an actual design recipe i could follow.
For example, coming back to the example of using fire, barrel of water, and finally smoke to sneak across some guard, that's not necessarily a bunch of systems which interact with each other, emerging a rich set of combinatorial options.
To me that's just one system: A physics simulation of rigid bodies and fluid, also modeling chemical reactions somewhat realistically. The option for the player to utilize the smoke is thus just what physics simulation adds to the table by default. And to describe this, i do not need the new term ‘systemic’ i have just learned.

To give the philosophy a purpose, i would need examples which work for my genre of interest - action games.
Maybe Deus Ex is an example. I don't remember it well, but when it came out, i liked the promise of ‘you can play the whole game without killing anybody’. There must be a lot of options was my expectation from the promise. But then, when i've played it, it was like a typical FPS, and i've killed any enemy as usual. It worked, so why should i spend time on figuring out complicated and abstract stuff like various electronic gadgets, augmentations, etc. There was no need to explore those options, and they did not feel accessible enough to make the exploration fun. The game was not a disappointment btw, it was good.
But it is maybe a good example for the primary problem i see: The interactions of various systems are not obvious, do not feel useful or profitable, and thus i could just make a dumb FPS for the same fun and invest my time on better enemy behavior, level design, and what not.

I do not like this final conclusion. It feels like resignation.
I hope i could come up with better ‘systems’, which work for me, and increase the fun by creativity, by a noticeable amount.
But so far i have not seen them in any game, which is a dilemma.

Maybe the new Prey is another example? People say it's the best immersive sim ever. So i tried to play it two times.
But it's boring AF. The enemies suck and it's no fun to fight them. And i can't find anything i could do in this game, which i could not already do in Quake, more or less. It's just bad in every sense imo.

So, systemic games look great on paper.
But in practice we have not figured it out at all yet.

…words of a dumb action games player ; )

aigan said:
It's often related to enemy encounters, which stems from the limitations of NPC interactions in the games of today.

Talking about NPC limitations, i also think that's a problem holding games back. Actually the largest problem, eventually.

But i come to slightly different reason about it: Static animation.
They can't do much, because they just play back static animations all the time. Adjusting footsepts with some basic IK does not fix anything.

Working on robotic ragdolls, i think i can break this limitation.
And then - trara… they will be able to do anything, and they will just do, and it will make my game oh so much better! \:D/

You think in the direction of richer conversation options to break limitations.

But if we are honest, and introduce a little bit of doubts to pull us back down to planet earth,
we are probably both just naive, and our stuff won't work as well as hoped at all.

However, that's imo just how gamedev works. So let's continue… : )

aigan said:
Do a lot of sneaking and get annoyed about being invisible to the enemy even when parts are clearly sticking out from the create I'm hiding behind. There are a lot of concessions made in stealth gameplay to make it less frustrating, and more fun. Doing any movement would in reality make you much more visible in tall grass, but I guess the thinking is that they don't want to discourage repositioning.

Personally i do not want to nitpick about an elbow popping out and the enemy can still not see me.
But the truth is: Stealth gameplay in action games does not work at all for me, because the generous help from the devs makes the whole mechanic totally unpredictable to me.

The Frictional games like Penumbra or Amnesia did it the best imo. It's a primary mechanic in those games, and it works. But without the horror, it would break down as well.

So yeah, that's really something i want to improve. Because like you, i want to avoid the twitchy and chaotic action. I want it steady and slow, and i want to feel safe and secure. I want to control the situation.

To make this possible, we need to address the oversight problem, which is again the cause of all evil.
I have an idea how to do this, and a first prototype seems promising. But i would loose shooting as the primary mechanic, and first person perspective as well.
What i have is like a mix of 3rd person camera and top down games which clip geometry occluding the player avatar. Promising, but sacrificing the immersion of first person feels hard to me.

aigan said:
It depends on what type of game you want to do, but I’d prefer higher stakes. Make the stealth more realistic and have fewer guards. Then, when you get caught

Hehe, i see my game design ideas never go very far. For example, i never thought about what should happen when getting caught. :D

Zak McCracken was nice. Getting to prison, removing all verbs to enforce a relaxing break.
That's a penalty i can live with. ; )

Btw, the definition of action adventure to me remains Flashback. Maybe the last 2D game i had played before the rise of 3D has erased all memories on other games for a decade or two.
Some years ago i have tried the game Shadow Complex, just out of boredom. Ik's similar to Flashback or maybe Metroid i guess. And it blew me away. This made me aware of the oversight problem.

@joej said:

To model some intelligence, NPCs should build up a mental model of potential future events forming a graph of causes and effects, then we could use greedy algorithms similar to A* to form a strategy and behavior. The problem here is clearly how to get those potential events, not the simple optimization problem to extract the strategy. I have not thought much about this yet, but interestingly my thoughts focus on the future (action game), while yours focus more on the past (NPC memory to achieve story).
We surely need to do both, probably using only one system. Past events could become persistent nodes in a graph representing memory. But there might be a need to forget things to keep it bound in size.

And i've called my ‘fun maker’ a ‘movie director’. : )

It’s worlds within worlds. Past, present and future. My model of your model of my model of your model. Also used for giving feedback of what is going to happen if you take certain actions, for a game with countless models that can hook into everything. No hand-written branching narrative, but lots of branching worlds, if only for different groups and persons beliefs of the past or plans for the future. Doing it hierarchical and flyweight are ways to keep it manageable and lightweight.

Likely our ideas are pretty common and there is a trend to get there.
But we may be late with some implementation ideas. Machine Learning will likely rush ahead, leaving us in the dust. ; )
I do however see potential advantages on precise control and performance, which is something important for games. And some people should keep working on ML alternatives anyway, so we can combine both in the end for best results.

Well. Short answer is that the right brain LLMs need to be combined with the left brain logic. The latest systems are using Q* and multi-agent for handling different perspectives. The structure will still have to be built. But its a good thing that there is machine learning for handling the categorization and pattern matchnings. Both for finding relevant threads, habits, templates, paths, plans, and in presentation for wordings, expressions, animations and speech. You can see more in https://blog.jonas.liljegren.org/surviving-the-burning-shores/#:~:text=Artificial%20Intelligence that also links to sections where I talk about it. … My incredibly invisible way to talk about the nature of AGI, buried in a Horizon video. ​

JoeJ said:

Physics simulators or AI rarely have any level of detail mechanism currently. It's just on or off, but can not approximate stuff at distance for lower runtime costs.

The systems of a systemic open world game would fall into the same category. So they must be designed from ground up for a moving boundary which just turns them off beyond. It's easy to underestimate what this means, in case those systems are meant to be global.

aigan said:
I'm going to write an article about the hierarchical world partitioning that can be used for simulations and procedural generation.

Please don't. : ) This problem is well researched.

It’s not for anything related to graphics or animation. This is for uprez and downrez the simulation for actors, the economy and world at large. It has more in common with the Victoria 3 Pops system.

The question is, if you put one stone on top of another. How long would you expect the stone to still be there when you come back? When can the fold in the specific stone back to the general stone density of the region?

Or for specific people that you recently met. If you turn away from them, they should still be in the same spot. But after a while, they could be in that general region. The movement will be downrezed back to the daily routine.

aigan said:
For a procedural generated world, locations can be generated based on what best fits the story and the player.

I could not do it. Generating scenery at runtime on the client would take hours.
Quality standards of AAA games impose the same limitations, mostly.

I mean, yes - we can build some houses from modular building blocks, implemented like base building in many games, or use some voxel engine, but that's low quality content then.

The thing I would myself think of attempting to do would not go beyond something like Rimworld. But I hope to play systemic story AAA first/third person games someday. For that, you should target a 2035 game system.

As long as the focus is just on graphics, it will eat up all the capacity, and then there will be the animations. That can go on for another 2 console generations. Multiplayer will half the fidelity. Systemic games, like Skyrim, will also half it. That is the price for making every object dynamic.

But the type of games and gameplay I like to see will not require systems to work at a per frame level. I will usually not run in new environments. The game will have 20 minutes or so to create the new location. Even so, if it actually tried to do something good, it could just ask for the time to work it out overnight. I could give it 10 hours if that was required. Or it could ask the cloud to help out to prepare something.

I saw the demo for Unreal 5 procedural generation adjusting the environment in seconds. Don’t think its an unsolvable problem. Especially if the world interactivity is valued enough, so that players understand the price of the larger data structures.

But besides that. I always thought that there could be graphical systems for baking geometry. If somebody put a stone on the ground, it could re-generate the geometry and shaders to make it as light-weight as possible, being a part of the ground. If you reach for the stone, it could swap in the older more detailed version.

My ideas about the hierarchical simulation system is also for generating new parts of the world on demand including passage of time from erosion to individual dresser drawers. The general idea is that most of the world is simulated, including passage of time, on a higher level for larger chunks of space. When you get closer, but still beyond the sight horizon, it will prepare the next level.

I started out with the one meter cubes, where a person occupies two cubes in height and can reach one cube over. This was for the original idea of a voxel-based world. But it can still be used for the simulation and generation regardless of the graphics actually used. With a factor of 16, there will be 7 abstraction levels. (1) People, (2) House, (3) Hill, (4) Mountain, (5) Run, (6) Region, (7) World. And there will be functions for up- and downrez between each level. And the simulation can run on a corresponding timeframe, from seconds on one end to decades on the other. This is for an actual earth-scale simulation. Most game worlds are super-compacted where countries have the size of what rather would be the size of cities.

Primary problem: How can the game figure out what the player wants and likes, or what's his fears?

I wonder if there were any attempts into this direction, avoiding a need for - ugh - neural sensors or brain interfaces.

Many games adapt in some manner to the individual player. The difficulty is adjusted. The amount of health and ammo is adjusted from what is needed. Games Like MGS5 will make the enemies adapt to the playstyle.

A systemic game can learn about the player by remembering their choices. The story genre will adapt to the player. If the player shoots at everything, the game can adjust to make it a lesser deal and not make every single NPC cry their hearts out. The playstyle will determine if the story is a comedy, drama, thriller, horror, action, romance, and so on. The topic of the story will be the thing the player interacts with. What type of people do they interact with, which places do they visit and what story do they continue on? The story will build on the thing the player does, like an improv yes-and methodology. Where do the conversations go? Since the game will offer up themes based on universal life experiences, it will see the things that the player finds meaningful. Even if the player just screws around, it will do it in their own way.

By just playing the game, the player will adjust the level of complexity, the length of the campaign, the type of relationships, the difficulty, the topics and the themes. They choose if they will go to “the dungeon of doom” or the “spring blossom festival”. They have declared their preference for the level of tension, tone, and so on.

In addition to that, they can also adjust the game directly through configuration menus. They can choose how peaceful or dangerous the world is, how complicated the stories will be, what type of conflicts there will be, how sorrowful or lighthearted the events will be, how much grind there will be, the degree of uncertainty, the level of guidance, how heavy the setbacks will be, and more. They could select genre and topics and starting scenarios. They can opt out of specific topics.

If you really wanted, you could expand this by allowing the game to read your social media or diary or whatever else. But I think it would be able to get to know you pretty well just by how you play the game. Even half a second of hesitation when seeing some certain new thing in the game could tell a lot about you. Like a tell-tale game, it's a giant psychology-test.

But let's say we get there, and the player can do anything he wants. To prevent boredom, the game also keeps him busy with constructing new enjoyable challenges and problems for him to solve.

Are we done then? No more need for new games, since they can't be any better?
We can just go home, play the game ourselves, turning real world off and enjoying a cozy time in the matrix, until humanity just dies out finally?

Hehe, idk. But thinking about this, i rather conclude it's about the path to get there, but not about the final goal. It's more about the promise and raising expectations, than about the actual achievement, maybe.
I prefer to keep selling snake oil rather than the blue pill. :D

Being better than any other type of media is not the same as “perfect”. And it's also a general statement, so it doesn’t have to hold for every person, or for the person's whole life. It's just a statement of, in average, the games impact on people's life, in what way they feel that it gives them something valuable.

And for every person, it’s a moving target. Like a great piece of literature. It will have an impact on them, and they will absorb it. But then, they will move on. The next thing could be something more within the game, or something completely different.

I would hope the game could give ideas of how to live better together on the planet.

aigan said:
It’s not for anything related to graphics or animation. This is for uprez and downrez the simulation for actors, the economy and world at large. It has more in common with the Victoria 3 Pops system.

Oh, that's something i have never read about. Surely interesting then.

aigan said:
The question is, if you put one stone on top of another. How long would you expect the stone to still be there when you come back?

A better example: Your game has destruction, and you destroy an entire building.
How do you feel if you came back and it's restored to its initial intact state?

Surely one reason we do not see serious destruction features in open world games.
(Also one reason why static worlds as actually a useful limitation. Once we lift it, we could no longer tell minimum storage space requirements.)

aigan said:
you should target a 2035 game system.

Oh no. Please god, let me get done with my crap before that. Not another 10 years. /:O\
You don't look younger than i. We might be dead til then… ; )

aigan said:
That is the price for making every object dynamic.

Dynamic objects cause no cost, as long as they do not move.

New console generations also no longer show massive increases in power. Speculated PS5 Pro update is quite subtle.
It also looks like most chip area they can still win from smaller circuits will be used only for AI acceleration.
At this point, i do no longer count on free lunch from newer, more powerful HW. The bit we still get is even compensated from low power devices like Switch and Steam Deck.
So we're stuck. Some fools are willing to pay premium for a 4090 for a real upgrade, but it's not enough worth making games for that.
MS is already bragging about the largest jump ever for their next gen, but if believe this at all, i guess it's a premium priced top model competing ridiculous enthusiast gaming PCs.
So the HW of 2035 wont be much more powerful than the HW we see already now, i think.
But i also don't need any further HW progress. Current HW is already more powerful than needed.
It's just that they already replace it before we can max it out at all, showing an increasing conflict of interest between HW and SW industry.

aigan said:
The game will have 20 minutes or so to create the new location.

Well, with increasing core counts that's possible.
But i still think we better invest the cycles into richer simulations, which is expensive.
And we want to create offline still, because only then we have art control as content creators.

It's artwork. You can not do it procedurally. If you do, it is no longer art. Or if you do it with generative AI, it's no longer art either. You accept a decline in quality, but increased dynamics isn't worth that price. At least mot in general.

Let's compare this with music. You want a game that creates itself while we play it.
That's like musicians improvising. It can be good, and for many people it's all they want.
But most people want a piece of art. Something remarkable. They want catchy earwigs, Beatles songs, Greatest Hits albums, No.1 Hits.
This requires ‘offline’ composition, which is not possible to achieve ‘at runtime’ from improvisation. Can't be done.
Computers, or computer programs, can't become artists.

Now it shows AI is pretty good at recycling stolen artworks of former artists, so maybe your vision becomes true. But personally i don't want it.

aigan said:
The topic of the story will be the thing the player interacts with. What type of people do they interact with, which places do they visit and what story do they continue on? The story will build on the thing the player does, like an improv yes-and methodology

1974: People can choose from Beatles or Led Zeppelin, or anything else. They choose what they personally like, and are happy with it. Which is no wonder, because those artists have achieved one masterpiece after another.

2024: People consume background music at a flatrate from spotify. There are playlists and channels serving any personal preference. But there is no more value in modern music. Composers can no longer come up with catchy earwig melodies. One song is like the other. The music industry is pretty much dead at this point. So i still listen the stuff from the 70's, often predating my time of birth.

We are facing the same direction with games. Currently it looks really bad. The last AAA game i have actually liked was Doom Eternal.
The games i still like are all indie games, made by small teams of 10 people at most.
In those games i still notice the contribution from individuals. Humans actually, not companies trying to unite the works of 400 unnamed artists. And that's what i want. That's what inspires me. That's what gives games, or artwork in general, it's value.

Notice my point here still applies to entirely procedural games like No Mans Sky, or to systemic games like NetHack or Dwarf Fortress. Those games a clearly the work of creative individuals. It's a manifestation of their ideas as living humans.
And the games industry is a creative industry. Automation is a tool for us, but not our destiny.
At least i hope so.

aigan said:
Even so, if it actually tried to do something good, it could just ask for the time to work it out overnight. I could give it 10 hours if that was required. Or it could ask the cloud to help out to prepare something.

No, you can't do that. It is a waste of energy. It's simply irresponsible to propose this. But due to your enthusiasm, i'll let it slip. ; )

aigan said:
I saw the demo for Unreal 5 procedural generation adjusting the environment in seconds. Don’t think its an unsolvable problem.

That's what i meant with ‘building houses from modular building blocks’, because that;s all EU5 can do.

The limitations are not obvious. The whole industry has converged on this practice, so we don't know what we miss.
The only counter example we have is Rage, which used unique textures everywhere. So no spot looked like the other, and there was no repetition.
Remember the first demo of UE5, showing a cave, many detailed rocks, and detailed ancient ruins.
All of this was composed from a finite set of modular models, each having many instances in the scene. It's a trick to achieve high detail from using many copies of the same model. Just like Super Mario duplicates the same block of grass or rock a thousand times in a level.
So the scene needs to be composed from those modular building blocks, and yes - we can move them around, or generate procedural scenery from it at runtime.
But to make such composition practical, models have to match each other. Walls need to fit doors geometrically, and worse: All rocks need to have the same color, otherwise you notice the trick pretty quickly.
As a result, even if we place the instances irregular but not on a grid, the scene becomes uniform and repetitive, especially once we want larger scenes, like open worlds. The old problem of video games.
That's also why Epic always only shows small and handcrafted scenes to impress, but no open worlds.
For that, they have to fall back to height maps, which lack detail compared to the scanned rock models.
Now they have added displacement mapping, so they can make the height maps detailed too, but the displacement maps are again made from tiling textures, bringing back the same limitations just at texture level.

Also, and that's what really matters: Composition from modular building blocks rules out the modeling on all frequencies. We are restricted to just one frequency, given by the size of the building block.
In nature, you see cause and effects at all frequencies. Erosion forms tendrils at all scales, and they are all connected to make sense, for example. Composition from building blocks rules out such connections, causign uncanny and unnatural appearance. Subconsciously, something feels missing. It's artificial. Basically the graphics analogue to ‘lacking choice’ or ‘game enforces the same outcome ignoring my former choices’.

The solution is to use a technology that enables unique detail everywhere, like Rage did, but only for textures.
I work on this. Not because i believe in unique detail to be so important, but rather because my path coincidentally brought me there. I work primarily on realtime GI, which requires global parametrization and hierarchical scene represtation fro LOD. Using the same system also for the visible geometry, i can basically achieve unique detail for little extra work.
But it's hard. Now you know where my past 9 years got lost. ; )

Once we arrive at unique detail everywhere, things like procedural generation and destruction at runtime just become harder as well, eventually.

But i live in my house for 20 yeras now, and it still looks the same. The trees outside - still the same. Neighbors houses - still teh same. The mountain in background - has not changed since thousands of years. But i can still see the causes and effects of the landslides happened form it back then. I can see wher all the rocks in the woods here came from. They tumbled down from that mountain, when the landslides did happen.
But i can not see such things in games.
Not yet, at least… ; )

It's ofc. a matter of compromise and priorities, which vary from game to game. There will always be runtime procedural worlds in games. But if you can, avoid making it a requirement for your upcoming AI middleware.

aigan said:
If somebody put a stone on the ground, it could re-generate the geometry and shaders to make it as light-weight as possible, being a part of the ground.

That's exactly what i do. The rock merges with the ground and there are no interior overlaps of redundant geometry. Output requires geometry at multiple levels of detail, collision geometry for physics, AI geometry like navmesh. And ofc. textures. I can have seamless displacement mapping everywhere. UE5 is surely still restricted to just planar patches, like all current tech. (Displacement mapping never became widely used because of this restriction coming from seams in texture space.)
If we want optimized data, and some amount of it, this means offline is the only way.
And it's many years of work ofc. Would require at least a small team of devs, actually.

Im still unsure if i'll use triangles at all for rendering, btw. Point splatting becomes attractive at high detail.
It can be either faster or higher quailty than triangles. I'll investigate…

Another problem is HW raytracing, which currently does not work for any fine grained LOD solution due to short sighted API design. It's currently useless if we look forwards, sadly. Epic can't trace its Nanite stuff either. They have to fall back to low poly proxies to put a ‘RTX on’ sticker on their stuff.
Big reason why i lost confidence in the HW industry. NV is a cancer, attempting to dictate progress, while understanding nothing about games. I hope x86 APU is the future for games, pushing them out of the market. They should focus on AI and leave us alone. But Qualcomms ARM CPU seems very promising, and if ARM becomes widespread, which it should, Jensens high price nonsense dictatorship will stay relevant, i'm afraid.

aigan said:
When you get closer, but still beyond the sight horizon, it will prepare the next level.

It's impossible to not underestimate what this means. Stuff at a distance feels small and thus easy to handle. But the problem is: It's a LOT of stuff. So you win nothing from hierarchical processing. It's the least you can do, but not a magical problem solver.
In this sense, our brain really tricks us with it's own reduction mechanisms, suppressing a sense of the actual quantity.
Ignore optimistic assumptions and do the math instead.

aigan said:
But it can still be used for the simulation and generation regardless of the graphics actually used.

Unlikely. It depends on what you simulate ofc., but if its' NPCs, the scale which matters to them can't be represented with 4x4m cubes for example. Pathways merge, doorways close, obstacles disappear. And that's just the very obvious examples, but the real problems will be in the details.
If the world is static, we can reduce it much better by pre-computing a low detail representations. We can preserve all paths and doors if needed.

But i don't think this should be needed either. I would turn the whole simulation into an abstract model driven by probability. NPCs just teleport around using estimated durations, change their data eventually. No spatial representation at all.
But there sure is a distance where the player can still see them, while there is no potential effect on each other. That's some work. I can disable physics sim for my ragdolls for example. It's still driven by balancing controllers, but the output is just procedural animation then. Still needs expensive collision detection though, but a capsule should do.
After that it becomes really difficult to avoid collision detection costs. Just fade them out smoothly to prevent popping, and good is i would say.
NPCs which are too distant to be visible likely need no further simulation, beside some single digit exceptions maybe. They could update their state by probability once they become close again to catch up.

aigan said:
Most game worlds are super-compacted where countries have the size of what rather would be the size of cities.

Yeah, luckily. Larger would not even be fun.

aigan said:
By just playing the game, the player will adjust the level of complexity, the length of the campaign, the type of relationships, the difficulty, the topics and the themes.

Yes, but then i'm not sure if the player really like the outcome.

Like, many years ago, i have clicked on a YT video of a guy reviewing Lego models.
And now i'm still haunted by that guy. He still pops up. Damn, i have clicked at it just once! I'm grown out of Lego!

I could give many more examples. It does not really work. Observing choices does not really tell you much about the player.

Contrary, if the game has no choice at all, it's the player doing his choice while selecting game A over games B,C,D. The player may still regret his choice, though.

But still, it's probably very hard to make a player adaptive game X to outsell games A,B,C or D, each targeting specific niches. And even while doing X, you may still almost certainly fail to give me the exact choice that i would want to do. So for me not much seems to change. I still have to pick the choice i dislike the least, but you have to work much harder on making the choices work.

Thinking of it, this is my main concern:

So for me not much seems to change. I still have to pick the choice i dislike the least

I'm serious about that. No chance i would notice the improvement. I play just once and would not notice.

So i think, maybe you should enforce me to notice. Maybe you should make the debug GUI i had mentioned a mechanic.
Could be as simple as letting the player die often, so he can select different choices next time.
Yes. make suicide a mechanic! \:D/
No, im serious. I think something like this is needed. Timetravel, or going back in time using some options menu to avoid the superficial, etc.
I wont agree with the choices you'll give me, but i might enjoy to explore them.