Player-Centered AI for Automatic Game Personalization: Open Problems

Jichen Zhu,Santiago Ontañón

DOI: https://doi.org/10.48550/arXiv.2102.07548

2021-02-15

Abstract:Computer games represent an ideal research domain for the next generation of personalized digital applications. This paper presents a player-centered framework of AI for game personalization, complementary to the commonly used system-centered approaches. Built on the Structure of Actions theory, the paper maps out the current landscape of game personalization research and identifies eight open problems that need further investigation. These problems require deep collaboration between technological advancement and player experience design.

Artificial Intelligence,Human-Computer Interaction

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is: **How to achieve automatic personalization of games through player - centered artificial intelligence (AI) technology to better meet the needs and preferences of players**. Specifically, the author points out eight open - ended problems in current game personalization research, which require further research and technological progress to be solved.

### Main Objectives of the Paper

1. **Propose a Player - Centered Framework**: Different from the traditional system - centered approach, the paper proposes a framework based on player needs and behaviors, emphasizing the combination of technology and player experience design.
2. **Identify Open - Ended Problems**: By analyzing the cognitive processes of players in games, the paper identifies eight problems that require further research, covering multiple aspects from player intention recognition to personalized content evaluation.

### Eight Open - Ended Problems

1. **Improve Modeling Accuracy (OP1)**:
- **Problem Description**: How to model the individual characteristics of players more accurately, especially when the amount of data is limited or difficult to label.
- **Challenge**: Obtaining sufficient individual player data and accurately labeling it is a difficult problem.

2. **Coordinate Theory - Driven and Data - Driven Models (OP2)**:
- **Problem Description**: How to coordinate the differences between existing theory - based (such as psychological theory) and data - based methods.
- **Challenge**: When the results of the two are inconsistent, it is not clear which side should be trusted.

3. **Reduce System Bias towards Certain Choices (OP3)**:
- **Problem Description**: How to ensure that players' choices can truly reflect their intentions and preferences, rather than being influenced by the in - game reward mechanism.
- **Challenge**: Players may choose options that help them win rather than truly represent their preferences.

4. **Provide Meaningful Choices for Players (OP4)**:
- **Problem Description**: How to design choices so that players' choices can lead to significant changes and avoid decision fatigue.
- **Challenge**: It is necessary to balance the quantity and quality of choices to ensure that the choices are both meaningful and not excessive.

5. **Increase the Controllability of Content Generation (OP5)**:
- **Problem Description**: How to algorithmically specify the characteristics of the required content and guide the generation algorithm to generate this content.
- **Challenge**: Current methods have problems of uncontrollability and quality when generating specific content.

6. **Evaluate the Quality of Generated Content (OP6)**:
- **Problem Description**: How to evaluate the quality of automatically generated content and avoid "catastrophic failure".
- **Challenge**: The existing automated evaluation metrics have insufficient correlation with human evaluation.

7. **Establish an Appropriate Personalized Psychological Model (OP7)**:
- **Problem Description**: How to help players understand the personalized mechanism of the system and form a correct psychological model.
- **Challenge**: An appropriate balance needs to be found between transparency and user understanding.

8. **Coordinate Players' Multiple Goals (OP8)**:
- **Problem Description**: How to coordinate among multiple player goals and ensure that personalized adaptation can meet different needs.
- **Challenge**: Different players may have different goals, and how to comprehensively consider these goals is a complex problem.

### Summary

This paper, by introducing the theory of players' cognitive processes, proposes a brand - new perspective to examine the current research status of game - personalization AI and identifies eight key open - ended problems. These problems not only involve technical - level challenges, but also involve how to better understand players' needs and behaviors, thereby promoting the development of player - centered game - personalization technology.