Privacy – large amounts of personal data collection give rise to privacy issues.

Data protection – data which is collected, stored and used is vulnerable to data protection issues and data breaches.

Autonomy – automated decision systems can reduce an individual’s autonomy in decision making.

Discrimination –  bias in data, design, development and deployment pose risks of discrimination.

Informed consent – lack of transparency on the use of AI and its internal mechanism can bring about issues of informed consent.

Liability and redress – lack of transparency concerning when AI is being used and the limitations in EU Liability regimes makes it hard for patients to prove they’ve been harmed by an AI system generated decision and get access to effective compensation.

Key populations include the aging population, people living with disabilities, people experiencing mental health problems, ethnic and racial minorities, undocumented migrants, members of the LGBTIQ+ community and people with insecure residence status, among others. They often already experience barriers in access to healthcare and are at risk of the following harms:

Discriminatory data – often, key populations are under-represented in data which is used to train algorithms. In addition, the data usually reflects real-world patterns of discrimination.

Privacy breaches and misuse of data – privacy invasion is a greater risk for key populations, because information may be used to their detriment/disadvantage.

Misdiagnosis, underdiagnosis and erroneous recommendations – as a consequence of discriminatory data and flaws in usage of AI systems.

Unequal access to technological innovations – due to accessibility requirements, digital literacy, language barriers or financial barriers.

Unequal access to healthcare – existing inequalities in access to healthcare may be replicated and exacerbated.

Errors and mistakes – AI generated decisions can simply be wrong. In healthcare this can have serious implications, including wrongful diagnosis or treatment.

Dehumanization – AI systems can never be empathic or truly caring. If they are used for caregiving this could replace human contact can lead to dehumanization.

Quick fixes – technological solutions can be seen as quick fixes for deep-rooted societal problems. It can then become one’s personal responsibility to ‘fix’ their health with tech applications. This can further lead to an overemphasis on efficiency and time savings instead of patient needs and satisfaction.

One size fits all – a personal approach should always remain pivotal in healthcare as needs, wishes and preferences differ per person. Mass use of AI technologies could undermine this value.