Should AI be allowed to make decisions on human life? 🔌

Artificial intelligence can be defined as any sort of system capable of performing tasks that typically require human intelligence, not just the generative AI the populous are accustomed to. The debate on whether such a structure should be allowed to make decisions on human life, which is any form of choice impacting us, revolves around the premise of accuracy, optimisation and reliability. Is the human experience better when surrendering the human choice on our life? Historically, such an argument has been met with ambivalent feelings. On one hand, its application in the medical field, in immigration, in aviation, has saved lives, reduced costs and shortened length processes significantly for the same, if not better, outcomes. Yet, such a proliferation has stagnated because of how alien AI is from humanity, its lack of accountability as well as its current inability to apply a moral framework. Thus, the question remains when coming to a conclusion: does AI need to be perfect to make decisions concerning our life?

To examine this case, an idea that initially needs resolving is the extent of application: should AI be used ubiquitously regarding decisions on human life, or only in exceptional circumstances? Such a question is pivotal to this argument, which can be exemplified by the aviation incident known as “The Miracle on the Hudson” in 2009. This case concerned the historical landing on the Hudson river after bird strikes destroyed both engines of the plane. The system on board analysed all the data available to synthesise a judgement to return back to where the plane departed, which simulations later suggested was possible. Yet, Captain Sully postulated an error within these models and algorithms: the inability to account for real world factors such as the time for pilots to process crises and the potential for marginal errors (after all this intelligence is built upon human input which is not flawless). This historical example highlights how such decisions that carry emotion, intense pressure and unexpected conditions require a human mediator because of AI’s inability to process unique factors. Whilst many could argue this to be an aberration because machines and systems such as autopilot are used on an everyday basis, this idea only holds ground when circumstances are invariable, unlike the scenario aforementioned. As a result, it could be initially argued that AI can be applied to many decisions that impact human life alongside the evaluation conducted by a human.

Such an assertion is reinforced through the application of AI in the healthcare field. AI is habitually used in artificial pacemakers, used as a diagnostic tool as well as an instrument for the removal of bureaucratic costs. The former use has helped treat life-threatening conditions such as arrhythmia, whilst also supporting the early identification of ailments such as cancer. The latter idea also demonstrated a significant economic burden which has long made the NHS inefficient and delayed the supply of healthcare to the most vulnerable. The extent of these ameliorations in a profession so significant to human life provides substantial evidence to support AI’s application on decisions impacting our life. Although the final decision is made by a doctor, suggesting human evaluation is necessary before execution, AI is crucial throughout the process. This is supported by its reliability in its conclusions as well as a tangible thought process which can be verified and critiqued, unlike the abstraction of human reasoning. 

Whilst the clarity and consistency cannot be matched, the problem that fosters concerns the lack of commonality with humans. This may largely arise because of the lack of accountability and moral framework (a set of principles that mediate between right and wrong) within its thought process. ‘Cold’ and ‘clinical’ justifies its working. Antithetically, one has security because a human perpetrator can be taken to court and held liable. But can the same thing be said about AI? Is it the institution behind the intelligence that is culpable? This clear uncertainty may be precisely the reason behind the fear of providing AI with this human decision. 

Furthermore, a greater concern with AI may be the capability of brutal information gathering. This has the potential for exploitation by parties, as demonstrated by the Cambridge Analytica data scandal, which harvested millions of personal data on Facebook. This possibility for corruption, combined with the ability for mass surveillance, is a crucial argument that limits the ubiquitous application of AI on human decision making. AI seems to replicate dominant ideologies during the time which can mitigate change and innovation, reducing the confidence when making impactful decisions. This seems to present Artificial intelligence as a double edged sword: when used altruistically, can reap a substantial reward, however, can also destroy. 

Whilst there is the possibility of risk, it is important to consider where AI can still has reward without cost. ‘Decisions on human life’ is a hypernym which suggests any choice that impacts the way life is conducted. There may be controversy in high risk applications, such as healthcare, because of its direct repercussions on morality, but can benefit on an operational level at lower risk. Contemporary critics have long commented on impartiality and lacklustre approaches in administrative roles such as immigration processing or prison risk assessments. Artificial Intelligence can exchange places with humans in such a process to inject reliability in its decision making process, which aforementioned, allows for its thought process to be analysed and refined. Whilst it may result in the redundancies in job posts, the innumerable benefits on citizens simply usurps the minimal, short-term economic instability that can occur. Even so, jobs can also arise oriented around AI, mitigating the extent of the economic problems. Whilst, on a metaphorical level, this may seem as if human life is a big red button with binary outcomes, the alternative - human decision making - is clouded with inconsistencies and on occasion, prejudice, which simply is not a sustainable approach. As a result, AI seems to have some role, if limited, on decisions impacting human life.

The Middle Ground:

Amalgamating this evidence may suggest a universal application of AI on human decisions seems improbable. Under the ‘Universal Declaration of Human Rights’ (UDHR) Article 3, ‘the right to life, liberty and security’ suggests that the end decision may ultimately be made by a human and no one else. However, in a world that is revolutionising with the usurpation of artificial intelligence, these mandates may be mandates partially outdated. In certain circumstances, such as within the medical profession, the mounting evidence for its application makes it impossible to disregard the copious benefits AI could provide if given the decision on human life. Additionally, neither humans nor AI are perfect: both make mistakes. It is the greater consistency and reliability AI provides, as well as the ability to continuously adapt, which inclines its use on these meaningful and charged decisions. Therefore, the most reasoned and potentially accepted conclusion could be providing scope for AI to make decisions on human life, but the end outcome mediated by a human.