[chongli]

I would say diplomacy is the most misunderstood feature of the series. Players constantly say they want a stronger AI that's smarter at diplomacy. But whenever they have built an AI like that, their play testers complained that it doesn't behave like a real world leader (too ruthless).

This experience led Soren Johnson (co-designer of Civ III and lead designer of Civ IV) to the realization that Civ AIs are supposed to "play to lose" [1].

[1] https://www.youtube.com/watch?v=IJcuQQ1eWWI

[Der_Einzige]

I am so tired of game designers/developers being so pathetically wrong about stuff like this. Modders have to CONSTANTLY fix these boneheaded, user hostile decisions in nearly every game. A lot of game developers are not the people actually loving/playing their games in the same way that the cello maker is usually not the cello player.

Even many popular mods fuck this up! DEI in Total War Rome 2 needs submods to make the AI play by the same rules as the player!!! This is top of the most subscribed list right now FOR A REASON!!! https://steamcommunity.com/sharedfiles/filedetails/?id=36258...

Make the AI play by the exact same rules as the player. Make a scaling AI difficulty slider which goes from "piss easy" to "insane grandmaster" but without cheats. It's not that hard to do this, the chess engine crowd figured it out back in 2001. FEAR figured it out in 2004. Game AI has straight up not improved and at many times gotten worse in the ensuing two decades.

[chongli]

It's not that hard to do this, the chess engine crowd figured it out back in 2001.

They really didn't. No one likes playing against weaker chess engines. They play perfectly like a higher-rated engine and then randomly make an obvious blunder. They don't play naturally like a human player of that rating.

The weaker AIs in Civ games do a far better job at "playing to lose" than low rated chess engines. It's not even close!

[stevage]

Yeah, the ability of chess engines to play like a human of a given rating is a pretty recent development.

And it's still a constant complaint that Stockfish suggests moves that no human could really play and follow up correctly.

[wewtyflakes]

That makes sense, but at the end of the day, it may be more fun to play around with opponents that act more relatedly. This could take the form of in-game/session-appropriate diplomatic responses that don't read like pre-canned text, or, having explanatory text for why the AI is acting perhaps in goofy ways (which comes up a lot).