Dynamic difficulty adjustment via AI: a trap for puzzle games

The appeal of dynamic difficulty

Dynamic Difficulty Adjustment (DDA), powered by AI tracking player performance, is a popular concept. The goal is noble: keep the player in the state of 'flow' by making the game easier if they struggle, and harder if they breeze through. In action games, adjusting enemy health behind the scenes can feel seamless. In logic puzzles, it is almost always disastrous.

Breaking the contract

A logic puzzle is a contract. The designer presents a set of immutable rules, and the player uses those rules to overcome an obstacle. If the game uses AI to silently remove an obstacle or alter a rule because the player is stuck, it breaks that contract. The player realizes they aren't solving a rigid system; they are being managed by an algorithm. The satisfaction of a hard-won victory evaporates.

Better alternatives

Instead of manipulating the puzzle behind the player's back, difficulty should be managed explicitly. Well-designed hint systems that respect the player's agency, branching paths that allow players to bypass difficult puzzles temporarily, and a meticulously crafted, static difficulty curve are far superior to opaque AI intervention.