Limitations
The cross-sectional design prevents causal inference. The observational design prevented causal inference. Self-perceived competence is distinct from objectively measured clinical performance. The experiments were simulations rather than l…
20 sources - 89 claims
The cross-sectional design prevents causal inference. The observational design prevented causal inference. Self-perceived competence is distinct from objectively measured clinical performance. The experiments were simulations rather than live deployments in hospital networks. The cross-sectional design prevents causal inference between behavioural intention and actual use. The study estimated potential social value rather than directly measured postinteraction behaviour. The study's generalisability is moderately constrained by its focus on University of Lagos affiliates in metropolitan Lagos. The evaluation is limited to 7B and 13B transformer-based LLMs. The study measured the number of ANC visits but did not evaluate the quality of ANC care. Self-report data may overestimate competence because of social desirability bias. OpenUp data were not linked to health records, so causal reductions in suicide attempts, deaths or service use could not be established. The evaluation is limited to FashionMNIST and does not establish scaling to larger natural-image datasets. The capacity measure had not been formally validated against observed behaviour. The backdoor proxy does not cover eve…