Compliance with HIPAA and other health regulations: NLP model must use real patient data during its processing of the notes, so it is important that the tool has safeguards for patient data and comply with health regulations. 

Ensure patient consent and trust for AI processing: the patient must feel comfortable with their data being processed by an AI assistant and give consent for it to do so. 

Address potential biases and misinformation: we will perform regular audits of the model’s output to identify and address potential biases systematically.

Mitigate effects of unrepresentative training sets: we will ensure the training data for Adelaide encompasses a wide range of demographics, clinical conditions, and geographical regions and regularly update the training datasets to include new medical research, guidelines, and diverse clinical cases.

We aim to further adress AI pitfalls through Mixed-Initiative Learning, which entails:

1. Collaborative interaction between AI and human experts

2. Joint contribution to diagnosis and treatment

3. Validation and guidance of AI outputs by human expertise

4. Involvement of all stakeholders to enhance transparency and trust

This is similar to Impetus at UCLA [3]. 

Address concerns about adding another application for healthcare professionals

Ensure user-friendliness and seamless integration with existing systems

Demonstrably improve efficiency and patient outcomes for acceptance