Researchers develop AIs to improve gamers’ dynamic difficulty adjustment

Researchers have devel­oped Arti­fi­cial Intel­li­gences (AIs) to improve gamers’ over­all expe­ri­ence by adjust­ing their dynam­ic difficulty.

Dynam­ic dif­fi­cul­ty adjust­ment (DDA), a recent devel­op­ment by Kore­an researchers, uses in-game data to pre­dict play­er emo­tions and adjusts the dif­fi­cul­ty lev­el in order to max­i­mize a gamer’s satisfaction.

Although dif­fi­cul­ty is a chal­leng­ing aspect to bal­ance in video games, doing it prop­er­ly is essen­tial to giv­ing gamers a sat­is­fy­ing experience.

The researchers’ work may help to bal­ance game com­plex­i­ty and enhance the appeal of games to dif­fer­ent sorts of gamers.

Gamers’ dynamic difficulty adjustment

Dynam­ic dif­fi­cul­ty adjust­ment (DDA) is a method for auto­mat­i­cal­ly alter­ing a game’s fea­tures, behav­iors, and sce­nar­ios in real-time depend­ing on the play­er’s skill so that the play­er does not get bored or annoyed whether the game is very easy or very difficult.

For instance, the game’s DDA agent may auto­mat­i­cal­ly raise the dif­fi­cul­ty if play­er per­for­mance exceeds the devel­op­er’s expec­ta­tions for a par­tic­u­lar dif­fi­cul­ty lev­el, rais­ing the chal­lenge for the gamer. This method is help­ful, but it has lim­i­ta­tions because it just con­sid­ers play­er per­for­mance, not how much plea­sure they are tru­ly having.

Gen­er­al­ly, games with dif­fi­cul­ty lev­els will run on a scale that includes some or all of the following:

  • Eas­i­er Than Easy,
  • Easy / Begin­ner / Novice,
  • Nor­mal / Medi­um / Stan­dard / Aver­age / Intermediate,
  • Hard / Expert / Difficult.

To help make its races more excit­ing and enter­tain­ing, regard­less of the skill lev­el of its play­ers, Mario Kart, for exam­ple, has includ­ed var­i­ous lev­els of dynam­ic dif­fi­cul­ty in near­ly every iteration.

Will AIs improve gamers’ dynamic difficulty adjustment?

A research team from the Gwangju Insti­tute of Sci­ence and Tech­nol­o­gy in Korea mod­i­fied the DDA approach in the study pub­lished in Expert Sys­tems With Applications.

They devel­oped DDA agents that adjust­ed the game’s dif­fi­cul­ty to opti­mize one of four dif­fer­ent aspects relat­ed to a play­er’s sat­is­fac­tion: chal­lenge, com­pe­tence, flow, and valence, as opposed to focus­ing on the play­er’s performance.

The DDA agents were trained using machine learn­ing, uti­liz­ing data col­lect­ed from real-world play­ers who par­tic­i­pat­ed in a fight­ing game against dif­fer­ent arti­fi­cial intel­li­gences (AIs) and then pro­vid­ed feedback.

Each DDA agent used both real-world game data and sim­u­lat­ed data to fine-tune the oppo­nent AI’s fight­ing strat­e­gy to enhance a par­tic­u­lar feel­ing, or ‘affec­tive state’, using the Monte-Car­lo tree search algorithm.

Asso­ciate Pro­fes­sor Kyung-Joong Kim, who led the study said that one advan­tage of the approach over oth­er emo­tion-cen­tered meth­ods is that it does not rely on exter­nal sen­sors, such as elec­troen­cephalog­ra­phy. “Once trained, our mod­el can esti­mate play­er states using in-game fea­tures only”, added Asso­ciate Prof. Kim.

Related: A way to “Detect Speech” from People’s brain

Through an exper­i­ment with 20 vol­un­teers, the researchers ver­i­fied that the pro­posed DDA agents could pro­duce AIs that improved the play­ers’ over­all expe­ri­ence, no mat­ter their pref­er­ence. This marks the first time that affec­tive states are incor­po­rat­ed direct­ly into DDA agents, which could be use­ful for com­mer­cial games.

Accord­ing to Asso­ciate Prof. Kim, com­mer­cial game com­pa­nies already have huge amounts of play­er data, which they can exploit to mod­el the play­ers and solve var­i­ous issues relat­ed to game bal­anc­ing using our approach. As men­tioned by the team, this tech­nique also has poten­tial for oth­er fields that can be ‘gam­i­fied,’ such as health­care, exer­cise, and education.

The researchers’ effort of devel­op­ing AIs could con­tribute to bal­anc­ing the dif­fi­cul­ty of games and mak­ing them more appeal­ing to all types of players.

Leave a Reply

Your email address will not be published.

Join our NewsletterDaily Glimple of Future

Our blog, "Daily Glimpse of Future", strives to make the future much clearer than it is today. Join our newsletter for free now.