Chatbot-based cognitive training: a signal of memory benefits in older adults without cognitive impairment

A program compared with no intervention

The researchers assigned participants to chatbot-based cognitive training, with 37 people, or a no-intervention group, with 35 people. The program involved 30 minutes a day, five days a week, for 12 weeks. These parameters describe the intervention studied, not a recommendation for use.

This randomized trial, described as single-blind, combined cognitive tests from the CANTAB battery with brain imaging at the beginning and end of the study. The groups were comparable at baseline. The abstract does not specify who was unaware of group assignments.

A memory finding, not a general cognitive improvement

The training group improved on the Paired Associates Learning test, a measure of episodic memory, with a P value of 0.04. However, the abstract provides neither the magnitude of this improvement nor its uncertainty interval. It is therefore impossible to determine whether the difference would be noticeable in everyday life.

On imaging, the researchers observed a smaller decline in fractional anisotropy in the tapetum of both hemispheres and the right posterior thalamic radiation. These findings persisted after correction for multiple comparisons. This measure provides information about white matter microstructure, but does not by itself demonstrate clinical protection of the brain.

Improvement on the memory test was also associated with changes in this measure in two brain pathways. This association does not prove that the changes observed on imaging explain the cognitive benefit.

The chatbot's specific role remains to be established

The lack of an active control group makes it impossible to distinguish the chatbot's specific effect from that of cognitive stimulation, engagement or expectations linked to the program. This trial therefore does not show that a chatbot outperforms another form of training.

The participants had no cognitive impairment, and the reported follow-up ends at 12 weeks. The findings do not establish effectiveness in people with dementia, a lasting benefit or prevention of cognitive decline.

What we can take away

This trial provides an encouraging signal on a memory measure and certain imaging measures. It is not enough to present chatbot-based training as a validated strategy for preventing cognitive decline.