Furthermore, the statistical reliability of these systems is subject to multiple threats, including overfitting, selection bias, spurious correlations, and algorithmic opacity ( Stiglic et al., 2020 ), which can lead to “hallucinations,” especially with large language models, which can have extremely significant impacts in high-stake setting such as medicine ( Bélisle-Pipon, 2024 ).
← all excerpts
Ethical issues raised by artificial intelligence and big data in population health: a scoping review.
1
—
—