How AI brain decoding models your thoughts without endless brain scans
Researchers showed that artificial networks can simulate how brains react to films, capturing distinct personal mental patterns without extra scanner visits.
Reading level
The full story with proper science words explained.
The trouble with giant magnets
What if an artificial system could watch a film and predict your exact brain reactions?
Reading thoughts usually requires functional magnetic resonance imaging (fMRI), a scanning method tracking blood flow to record neural activity. These scanners are huge and costly. They also require people to lie totally still inside a narrow tube. Because of this, testing lots of video clips is very hard.
Building a neural stand-in
To solve this problem, researchers tested a new route using AI brain decoding, which translates neural patterns into recognisable thoughts. First, they trained deep neural networks (DNNs) on human fMRI data. These computer models learned from scans taken while people watched movie scenes. In this way, they learned how specific scenes triggered unique brain reactions in each person.
Next, the team showed brand-new video clips to the trained models. The system created fake brain signals for these clips without scanning the human again. Software decoders then read mental states straight from these software predictions. The model matched real fMRI scans for more than 70% of 88 tested cognitive labels, which are descriptive tags for specific thoughts or concepts.
Why simulated minds matter
Because scanners are so awkward to use, fake brain reactions could help fields like neuromarketing, which studies consumer preferences. Instead of booking hours of scanner time, teams could test new adverts against a person's computer profile.
The method relies on computer predictions rather than live scanner data. It also failed to capture patterns for nearly 30% of the tested tags. A computer model cannot replace the first scanner session needed to train the network either.
What we still do not know
Researchers have not tested whether this setup works beyond movies. We do not know if it works with spoken sound, written text, or interactive tasks. It is also unclear if accuracy holds up over long periods or across diverse groups left out of the early training data.
Science words
- functional magnetic resonance imaging (fMRI)
- A brain scanning technique that tracks blood flow changes to measure neural activity.
- AI brain decoding
- The process of translating measured or predicted brain activity patterns into recognisable mental states or perceptions.
- deep neural networks (DNNs)
- Computational models inspired by biological neural networks used to learn complex patterns from data.
- cognitive labels
- Descriptive tags categorising specific thoughts, perceptions, or mental concepts associated with stimuli.
- neuromarketing
- The commercial application of neuroscience tools to study and predict consumer responses and preferences.
Check it yourself
This story is based on a real research paper in NeuroImage by Kawahata, Wang, Blanc et al. We write with AI help and check it against the paper, but the original is the final word.