70/100

Moderate evidence

Cite
More
EmailLinkedInX

Older adults rated exercise as less strenuous with an AI companion, a 94-person study finds

Study title: AI-Supported Companionship During Older Adults’ Physical Activity: A Comparative Experimental Study of Post-Exercise Perceived Exertion

Evidence type
Randomized controlled trial
Sample
n = 94
Sources
1 sourcePeer-reviewed
Published
2 min read

Study title: AI-Supported Companionship During Older Adults’ Physical Activity: A Comparative Experimental Study of Post-Exercise Perceived Exertion

Researchers randomly assigned 94 older adults to one of four conditions while they did 4 sets of 12 dumbbell curls with 2.5 kg weights: no companionship, human companionship, low-interaction AI or high-interaction AI. Companionship condition significantly affected post-exercise perceived exertion on the Borg 6–20 scale (F(3, 90) = 11.407, p < 0.001, η2 = 0.275).

Read the study abstract

Background/Objectives: Maintaining physical activity is important for functional independence in older adults. Perceived exertion may affect their exercise experience and participation. Social companionship may alleviate perceived exertion, and artificial intelligence (AI) offers a new form of companionship. This study aimed to compare human and AI companionship on post-exercise perceived exertion and examine the relationships among AI acceptance, social presence, and perceived exertion. Methods: Ninety-four old…

AI-Supported Companionship During Older Adults’ Physical Activity: A Comparative Experimental Study of Post-Exercise Perceived Exertion
"Pathologists looking into microscopes (1)" by Linda Bartlett, public domain.

The gain

In a randomized experiment of 94 older adults doing 4 sets of 12 dumbbell curls, those exercising with a low- or high-interaction AI companion reported significantly lower post-exercise exertion than those exercising alone (both p < 0.001).

In brief

Participants exercising alone rated the effort as harder than those with a human companion (p = 0.003) or either type of AI companion (both p < 0.001). The three companionship groups did not differ significantly from one another, so the more interactive AI was no better than the simpler AI or a person.

Within the AI groups, a stronger sense of another's presence was the only significant predictor of lower exertion (β = −0.323, p < 0.05), and perceived usefulness predicted that sense of presence (β = 0.326, p < 0.05). The authors call these analyses exploratory and say a larger sample is needed to confirm them.

Main points

  1. The study randomly assigned 94 older adults to no companionship, human companionship, low-interaction AI or high-interaction AI.
  2. Each participant did 4 sets of 12 dumbbell curls with 2.5 kg dumbbells, and effort was rated on the Borg 6–20 scale.
  3. Companionship condition had a significant effect on exertion ratings (F(3, 90) = 11.407, p < 0.001, η2 = 0.275).
  4. Exertion ratings were higher with no companionship than with human (p = 0.003), low-interaction AI (p < 0.001) or high-interaction AI (p < 0.001) companions.
  5. There were no significant differences among the three companionship groups, and high-interaction AI was not shown to beat low-interaction AI or human company.
  6. In the AI groups, perceived usefulness, sense of others' presence, emotional bonding and perceived interaction were each negatively correlated with exertion.
  7. Sense of others' presence was the only significant negative predictor of exertion (β = −0.323, p < 0.05); perceived usefulness positively predicted it (β = 0.326, p < 0.05).

The full story

The experiment compared four conditions among older adults, with the aim of testing whether AI could offer the same kind of social companionship that a person does during exercise. Maintaining physical activity matters for functional independence, and perceived exertion may shape whether older adults keep exercising.

Analyses used one-way ANOVA, Pearson correlations and regression. The AI groups additionally completed measures of AI acceptance and social presence, which allowed the authors to examine which perceptions went with lower exertion ratings.

The authors conclude that human and AI companionship were associated with lower immediate post-exercise exertion than none, and that findings support further investigation of AI exercise companions combining functional guidance with social interaction.

Sources

1 on file
  1. Healthcaredoi:10.3390/healthcare14223885
    Peer-reviewed2026-11-12

Talk it over

Three questions to start on, two more if you have time
Older adults exercising alone rated the effort as harder than those with a human companion or either AI companion, and the three companionship groups did not differ. What does that suggest about what the AI was actually adding?

Start with a question

Older adults rated exercise as less strenuous with an AI companion, a 94-person study findshttps://truvace.com/item/ai-supported-companionship-during-older-adults-physical-activity-a-comparative-e

  1. Older adults exercising alone rated the effort as harder than those with a human companion or either AI companion, and the three companionship groups did not differ. What does that suggest about what the AI was actually adding?

    Answer in the thread
  2. The study measured only immediate post-exercise exertion after 4 sets of 12 dumbbell curls with 2.5 kg weights. How far can that tell us whether older adults would keep exercising over weeks or months?

    Answer in the thread
  3. Which older adults might not benefit from an AI exercise companion, and who might be left out of a group of 94 volunteers doing dumbbell curls?

    Answer in the thread
If you have more time
  • The authors found that a stronger sense of another's presence predicted lower exertion in the AI groups and call this exploratory. How would you design a larger study to test it?

    Answer in the thread
  • If you were choosing an exercise companion for an older relative or neighbor, which would you pick among a person, a simple AI or an interactive AI, and what would drive the choice?

    Answer in the thread

What readers say