How AI Can Be Used to Make Games More Accessible?

The use of AI in the video game industry does not always mean a threat to developers

The use of artificial intelligence (AI) technology in the video game industry has sparked debate about the pros and cons. There are those who think AI will help produce better games but there are also those who worry that AI will take over their jobs.

This kind of polarization raises questions about the future prospects of AI in the gaming industry. Will AI have a very crucial role in the gaming industry? On the other hand, will the use of AI in game development continue to be blocked as it is currently?

The answer may depend on the goals of each party using AI. If the aim is good, the results may also be very profitable for many parties.

Last May, Google unveiled Project Gameface an AI-powered gaming mouse prototype designed for users with disabilities. Using Gameface, users can control the computer cursor with head movements or facial gestures.

The main inspiration behind the project is Lance Carr, a quadriplegic video game streamer who uses a head-tracking mouse to play. When a fire engulfs Lance’s residence and destroys his gaming setup Google tries to offer help by creating an affordable and easily customizable open-source alternative that leverages machine learning.

AI vs. AI machine learning

To understand how Gameface works, we need to understand the definition of AI and machine learning first. To Wired Laurence Moroney, one of the brains behind Gameface, explained the difference between the two: “AI is a concept. Machine learning is the technique you use to implement that concept.”

In other words, machine learning falls under the AI ​​umbrella, just like other implementations such as large language models (LLM). But unlike ChatGPT or Stable Diffusion which are iterative in nature, machine learning is designed to be able to learn and adapt without instructions, by drawing conclusions from patterns that can be read.

Laurence explains how Gameface involves multiple machine learning models at once. The first model can detect the location of faces in an image. When the face image has been identified, it is the second model’s turn to understand the location of the obvious points (eyes, nose, ears, and so on).

After that, another model will map and interpret the gesture from these points, before finally translating it into mouse input. The implementation is actually helping rather than replacing, and Laurence believes this is the best way to apply AI, namely to expand “our capacity to do things that were previously impossible to do.”

This sentiment doesn’t just apply to Gameface and its potential to make gaming more accessible. According to Laurence, AI can not only have a big impact on accessibility for players but also on the way developers create accessibility solutions.

AI to develop accessibility solutions

A number of developers have recently begun to have a similar understanding. Artem Koblov, a creative director at game studio Perelesoq, told Wired that he would like to see “more resources directed towards completing routine tasks rather than creative discovery.”

Thus, AI can help in time-consuming technical processes. With proper application, AI can create a more efficient, more permissive development cycle and help in the implementation of accessibility solutions, while giving developers more time to consider them.

“As a developer, you want to have as many tools as possible to help you make your job easier,” says Conor Bradley, creative director at Soft Leaf Studios. He gave examples of a number of AI implementations in accessibility that are currently available, such as real-time text-to-speechspeech-to-text generation, and speech and image recognition

According to Conor, over time more and more developers will take advantage of this series of AI tools to make their games more accessible.

Artem Koblov believes this can be developed further. He envisions AI being trained to create a fundamental accessibility framework that can be adapted and embedded into games.

“This kind of framework will adapt the visual, audio and interactive aspects of the game. In other words, small developers like us don’t need to do expensive research, develop unique solutions, or test iteration after iteration of solutions that we create ourselves,” explained Artem.

However, this should not be misconstrued to mean that human input is no longer needed. According to Conor Bradley, AI is full of potential, but AI is by no means a shortcut.

“You can’t say, ‘AI, make my game accessible!’ and suddenly you now have the most accessible game of the year. We need players, including those from the disabled and neurodiverse communities to test our games. Ultimately, it’s humans who will play your games, not machines,” explained Conor.

Again, when discussing the implementation of AI, we must have the mindset that AI is only an addition, not a replacement.

Another example of AI implementation to make games more accessible is Minecraft Access a mod created to make Minecraft easier to play for blind users or those with other visual impairments.

This mod involves many AI tools at once, including ChatGPT and Google Tensor Flow. “We hope AI can provide visual context for blind and low-vision players by providing information about the world according to needs and requests,” one member of the mod development team told Wired.

Challenges in implementing AI for accessibility solutions

The potential of AI will sound even more interesting when it doesn’t just improve accessibility, but also actively learns what players need. This will be very useful for wider applications in accessibility, especially considering how wide the disability spectrum is and how personal each player’s needs are.

Nevertheless, we still have to be able to control our expectations. The reason is, even though these AI implementations sound very promising, in fact there are still quite significant obstacles. The simplest example, Minecraft Access requires several different programs to function properly, and this is probably not something most users can or want to do.

Returning to discussing Gameface, Ben Green, as a gamer with a disability, expressed his concerns, especially regarding the diversity of data used to train the AI. According to him, Gameface may be able to recognize many faces, but he doubts this will also apply to those who have significant differences in their faces, such as people who must continue to use ventilators.

Responding to these kinds of concerns, Miguel de Andrés-Clavera, head of the Gameface development team, said that they had designed the system to be easy to customize. Users can basically choose which facial expressions they want to use to control the mouse and they can also adjust the intensity of each selected gesture.

Hearing these stories, it can be concluded that sentiment towards the use of AI in the video game industry does not always have to be negative. Just like in other industries, AI can also be misused here. But everything comes back to the abilities and desires of each person who uses AI.

With the right direction, AI can be used to make games more accessible to more players. Solutions that previously seemed impossible to create, will later become possible thanks to the help of AI. At that point, AI may seem more like a superpower to game developers than a threat.

Leave a Comment