Why do our eyes skip and reread words? AI may reveal how we read efficiently
Scientists let an artificial intelligence (AI) model work out the best way to read for itself and it ended up behaving strikingly like human readers. Researchers at Aalto University in Finland, the Hong Kong...
Scientists let an artificial intelligence (AI) model work out the best way to read for itself and it ended up behaving strikingly like human readers.
Researchers at Aalto University in Finland, the Hong Kong University of Science and Technology, City University of Hong Kong and the National University of Singapore used AI to better understand how people read.
“For the first time we’ve used AI methods to understand – not just mimic – how people read,” Antti Oulasvirta, a professor at Aalto University, said in a press release.
The team built the model around the idea that readers have limited time, memory and ability to take in visual information.
They then trained the AI on millions of texts to maximise how much it could understand within those constraints.
Even though no human gaze or comprehension data were used during learning to copy these human reading habits, researchers found that many of the same patterns emerged when they compared the model with real readers.
For instance, it spent less time on common or predictable words and was more likely to skip them, while difficult or ambiguous words and sentences prompted it to go backwards more often.
“Reading feels effortless, but your brain is constantly deciding where to look, what to skip, and when to backtrack, spending attention like a budget to maximise understanding,” said Shengdong Zhao, a professor at City University of Hong Kong.
Researchers call this “resource rationality”, meaning limited time and attention are directed where they are most useful for understanding the text.
The AI model adapted to time pressure in a similar way to 39 adults who participated in an experiment. When time was short, it prioritised covering more of the text. With more time, it spent more attention on difficult passages and reread more.
However, the results do not prove that the human brain performs exactly the same calculations as the AI, researchers caution.
Instead, the researchers say that the model offers a possible explanation for how different reading behaviours could emerge from the goal of understanding as much as possible with limited cognitive resources.
Towards more personalised reading
A better understanding of how memory shapes reading could eventually help create technologies that adapt text to individual readers and situations, such as customisable apps, according to the research team.
“Now we have this new understanding of something that’s so central to our lives, it’s just a matter of exploring all the possibilities,” said Oulasvirta.
One potential use is augmented reality (AR). The approach could be used in smart glasses to adjust the pace or layout of on-screen text according to a user’s needs and what they are doing.
“We could take the same source text, say, a convoluted piece of legal writing and with little effort produce versions that are more comprehensible for different readers,” said Oulasvirta.
The team plans to investigate whether the technology could help people with dyslexia or lower language proficiency. Researchers also envision adapting text in situations where attention is particularly important, such as presenting information to drivers without unnecessarily distracting them.
“We want to help users in real-time situations, for example, by designing text that helps drivers without distracting them,” Oulasvirta said.
For more on this story, watch the video in the media player above.
Video editor • Roselyne Min
