2022 Fiscal Year Final Research Report
Progress support programs that respect game players' intentions and values
| Project/Area Number |
20K12121
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 62040:Entertainment and game informatics-related
|
|---|
| Research Institution | Japan Advanced Institute of Science and Technology |
|---|
Principal Investigator |
Ikeda Kokolo 北陸先端科学技術大学院大学, 先端科学技術研究科, 教授 (80362416)
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Keywords | ゲーム / 楽しませるAI / 好みに寄り添うAI / 囲碁将棋 / スーパーマリオ / 価値観推定 / プレイヤーモデリング / 強化学習 |
|---|
| Outline of Final Research Achievements |
We aimed to develop techniques to model player behavior and preferences for multiple games, and to play them as opponents or friends accordingly, and to create problems for them.
As an example, for Super Mario, we proposed a reinforcement learning task that generates stages to maximize the stage desirability as a reward, and confirmed that it produces diverse and high-quality stages.
As another example, for Go and Shogi, we modeled moves that humans consider natural by combining supervised learning and reinforcement learning. We then proposed a technique for guiding the AI to play well with natural moves against beginners and intermediate players, and confirmed the effectiveness of the technique through experiments.
|
| Free Research Field |
ゲーム情報学
|
| Academic Significance and Societal Importance of the Research Achievements |
AIが人間から仕事を奪う敵なのか幸せを導く味方なのか，市井から研究者まで多くの議論がなされている．本研究は，人間を打ち負かすまたは人間が従わなければならない強いAIプレイヤではなく，人間によりそい，ひきたて，楽しませる暖かいAIプレイヤを作成することを目指し，いくつかのゲームにおいて目的を達成した．
ゲームはもともと楽しむために行うものであるが，いずれはゲーム以外のゲーム的実問題においても，人間の弱さや好みに寄り添ったAIの活用法が開発され，広まりAIとの適切な付き合い方できるようになると考える．
|
