The continuing development and deployment of AI technologies in diverse domains has more and more people interacting with AI-powered products in their everyday lives. We deal daily with AI assistants like Alexa or Google Home, while robot vacuums clean our floors, algorithms trade our stocks and route-planning apps optimize our trips.
Although most human-AI interaction is productivity-based, new research suggests that human gamers’ interactions with AI — “AI as Play” — can contribute to improvements both in user experience and human-computer interaction (HCI) in general.
Human-AI interactions in games are increasingly common, building on AlphaGo‘s mastery of the ancient board game Go and the success of AlphaStar against human pros in popular video game StarCraft. In the paper Player-AI Interaction: What Neural Network Games Reveal About AI as Play, researchers from Drexel University, Northeastern University and IT University Copenhagen explore how humans interact with AI in such contexts, with a focus on computer games.
The researchers chose 38 neural network games for their systematic review of player-AI interactions. They employed a two-phase qualitative analysis, considering player-AI interaction in neural network games and neural network games with general human-AI interaction guidelines.
In the first phase, Analyzing Player-AI Interaction in NN Games, the researchers examined how these games use neural networks and how the networks structure player-AI interactions, identifying the “overarching interaction metaphors and patterns” of how the neural networks are represented in a game’s user interface.
In the second phase, Analyzing NN Games With General Human-AI Interaction Guidelines, the team applied human-AI interaction design guidelines from various current productivity-based domains to their dataset. This provided insights on the design of human-AI interaction, how neural network games comply with contemporary design guidelines for human-AI interaction, and, more importantly, how neural network games differ from other AI-powered products under the design guidelines for human-AI interaction. Adapting these human-AI interaction guidelines to player-AI interaction helped identify the strengths and weaknesses of neural network games.
Based on their findings, the researchers propose the notion of “AI as play” as an alternative to the current paradigm of performance-centric human-AI interaction that can contribute to both the game design and HCI research communities.
The team also identified several limitations in their work, such as potentially overlooking some relevant neural network games and omitting failure-related human-AI interaction guidelines. They propose that future research could focus on failure in games and how it relates to player-AI interaction, and adopt a more fine-grained qualitative analysis.
The paper Player-AI Interaction: What Neural Network Games Reveal About AI as Play has been accepted by the ACM CHI Conference on Human Factors in Computing Systems (CHI2021), a premier international conference on human-computer interaction. The paper is available on arXiv.
Analyst: Yuqing Li | Editor: Michael Sarazen; Fangyu Cai
This report offers a look at how China has leveraged artificial intelligence technologies in the battle against COVID-19. It is also available on Amazon Kindle. Along with this report, we also introduced a database covering additional 1428 artificial intelligence solutions from 12 pandemic scenarios.
Click here to find more reports from us.
We know you don’t want to miss any news or research breakthroughs. Subscribe to our popular newsletter Synced Global AI Weekly to get weekly AI updates.