CAN LARGE MULTIMODAL MODELS INTERPRET HEALTH CARE INFOGRAPHICS IN DEMENTIA CAREGIVING MATERIALS?
2024

Can Large AI Models Understand Health Infographics for Dementia Care?

Sample size: 5 publication Evidence: moderate

Author Information

Author(s): Luo Zhimeng, Zou Ning, Xie Bo, He Daqing, Hilsabeck Robin, Aguirre Alyssa

Primary Institution: University of Pittsburgh

Hypothesis

Can large multimodal models interpret health care infographics effectively?

Conclusion

GPT-4V significantly outperformed other models in interpreting the completeness of health infographics for dementia caregivers.

Supporting Evidence

  • GPT-4V achieved an average accuracy of 99.0% across the infographics.
  • The average completeness for GPT-4V was 93.7%, significantly higher than other models.
  • Cohen’s kappa for coder agreement was 0.8205, indicating high reliability.

Takeaway

We tested some smart AI models to see if they could understand pictures and information meant for people taking care of dementia patients, and one model did the best job.

Methodology

Four large multimodal models were tested on 5 infographics, with their interpretations analyzed for accuracy and completeness by independent coders.

Statistical Information

P-Value

0.04

Statistical Significance

p=0.04

Digital Object Identifier (DOI)

10.1093/geroni/igae098.0913

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication