Summary Ethical & Legal challenges AI driven healthcare

As most of you know me personally, I like writing a short piece on AI and what I am currently motivated by. Giving a little more of a personal touch when it comes to my experiences in Healthcare. I followed for many years the classical medical school program, worked long hours in hospitals — weekend shifts, night shifts, I am sure we all know the drill. And then in Spain started working for Online Doctors (#NLP chatbot) style of work. We encountered some ethical and legal challenges here. Making diagnosis online without physically seeing the patient, good or bad idea? How do we deal with bias? How to we deal with our data collection? One extremely interesting observation I made was — patients are less ashamed to speak online about their body because they do not feel embarrassed.

The hype around artificial intelligence in 2021 has been immense. Machine learning is the most popular approach where computation systems learn from data and improve the performance à opening the black boxes. A part of machine learning is what it is know as Deep learning which employs artificial neural networks with multiple layers of larger datasets.

Trends: in the USA since Trump AI strategy shifted to free market-oriented approach, also known as the ‘hands off’ approach, the White House published draft guidance for regulation AI applications (2020) which included:
1. Public trust
2. Public participation
3. Scientific integrity & information quality
4. Risk assessment & management
5. Benefits & costs
6. Flexibility
7. Fairness & nondiscrimination
8. Disclosure & transparency
9. Safety & security
10. Interagency coordination

European AI Healthcare — is worth 20 Billion by 2020
— European AI Alliance published ethics guidelines (April 2019) known as ‘Trustworthy AI’ which identifies 7 key requirements to fulfill the order to be trustworthy.
1. Human Agency & oversight
2. Technical robustness & safety (@neurocat.ai)
3. Privacy & data governance
4. Transparency
5. Diversity, nondiscrimination & fairness
6. Environments & societal well-being
7. Accountability

ETHICAL CHALLENGES
To what extend do clinicians have to be responsibility to educate the patient around complexities of AI? What data inputs have been used? Machine Learning processes? Biases? How much transparency is needed here between patient care, health delivery and quality of the doctor’s work?

Safety & Transparency

Safety — is biggest challenge in AI healthcare — example the IBM Watson, was supposed to help physicians explore cancer treatment options, using #synthetic cancer cases. However, it failed to make the right diagnosis.

Potential of AI has two key ingredients
1) reliability and validity of datasets (data sharing)
2) transparency (shortcomings of software- data bias)

Algorithmic Fairness & Bias

Bring AI to remote areas! This is a wonderful idea, for remote areas to have access to top quality healthcare world-wide. We are making progress, but all Machine Learning systems are human trained algorithms; therefore, will only be as trustworthy effective and fair as the data that it is trained with. With bias here we mean ‘discrimination’ in patient files the skin color, disabilities, age, gender, to name a few, all need to be considered. Phenotype and genotype can lead to false diagnosis and therefore need to be considered carefully. ‘#Black-box’ algorithms — XAI is necessary when AI makes health recommendations, especially to detect biases. How does the AI reach the final decision (inference)? But instead we also need to consider is the AI safe and accurate for us to use in Healthcare.

Data privacy

Who has the ownership of data? The patients? From a survey on PubMed the public is uncomfortable with companies or the government selling patient data for profit => people emphasized to what extend can patients withdraw their own data? And when they have their data will there be someone who can explain it in a constructive manner.

I identified some key facts of USA and Europe and were we stand with AI currently.

As we see there are quite some differences between continents how to deal with the future of AI in Healthcare. I feel it is my responsibility to facilitate the ethical and technological advancement in AI and most importantly confluence data privacy, patient interest and an entrepreneurial spirit in a clear and effective way.

Currently I am working as a medical executive at Neurocat (https://www.neurocat.ai/ ) where I am dealing with my clients around the topic robustness and how to improve their AI in order to get certified.

In summary:

Main ethical challenges for AI in healthcare

1. Informed consent to use

2. Safety and transparency

3. Algorithmic fairness and biases

4. Data privacy

5 legal challenges

1. Safety and effectiveness

2. Liability

3. Data protection and privacy

4. Cybersecurity

5. Intellectual property law

(We need to create a system that is built on public trust!!!!)

My motivation is to create this trust by offering transparency and guide my readers through the ethical and technological challenges. Stay tuned on my profile for more advancements in AI.

Please reach out apu@neurocat.ai