For centuries, technology has helped human beings reduce physical effort. Machines lifted heavier loads, factories accelerated production, and computers processed information faster than any individual could. Artificial intelligence, however, marks a more intimate shift. We are no longer asking machines only to perform tasks, we are increasingly asking them to interpret situations, rank people, recommend actions and shape decisions.

A current controversy in Australia shows why this distinction matters:

The Integrated Assessment Tool (IAT), introduced in November 2025, helps classify elderly care needs and determines the eligibility and classification of services, including residential care and government-funded home support. The Australian government presents the tool as a way to improve consistency, but the way IAT addresses home support eligibility has become a central concern. 'The algorithm-based assessment tool is 'cruel' and 'inhumane', stripping away clinical expertise and leaving elderly people with inadequate support,' says aged care clinicians and carers. The government's IAT user guide states that aged-care assessments should rely on an assessor’s expert judgement, but there are only limited circumstances in which the decisions of the IAT can be overridden by the assessors.

In one reported case, a man living with late-stage motor neurone disease was denied a higher level of funding even after seeking a review. The controversy eventually reached Parliament, where the Australian Senate passed the Aged Care Amendment (Restoring Human Override for Aged Care Needs Assessments) Bill 2026 on July 2nd, despite the fact that the proposal still required approval from the House of Representatives. The bill was passed to restore assessor discretion and overturn computer-generated funding outcomes if an algorithm under-identifies elderly care needs.

This is not simply a debate about whether one administrative tool works well. It reveals a deeper political question: Who is actually judging?

A nurse may interview an elderly person, observe their vulnerability and understand the details of their daily life, but, when those observations are translated into boxes, scores and classifications, the final outcome can acquire an authority of its own. The professional remains present, yet their judgement becomes secondary. Human participation is reduced to feeding information into a system and approving what comes back.

Hannah Arendt helps us see the danger in this arrangement. Her concern with thoughtlessness was not a criticism of low intelligence, it was a warning about the surrender of reflection: the moment people stop examining what they are doing because a procedure, institution or authority appears to have already decided for them. For Arendt, the absence of thinking could produce an inability to judge precisely when judgement was most urgently required.

In an AI-assisted bureaucracy, “the system recommended it” can become the modern equivalent of “I was only following the rules.” Responsibility is not eliminated; it is dispersed until no one feels fully answerable.

Kant would describe the same problem as a loss of intellectual maturity. His call to use one’s own understanding was not a demand that we reject guidance. Human beings always depend on knowledge created by others. The real issue is whether we can evaluate that guidance and give reasons for acting upon it.

Assistance becomes dependence when we no longer feel capable of questioning an output—or when institutions discourage professionals from departing from it.

Aristotle adds another essential insight. Good judgement requires practical wisdom: the ability to recognise what a particular situation demands. Rules can guide us, but no set of rules can fully anticipate the complexity of human life. For Aristotle, what should be done in a particular case depends upon circumstances that cannot always be resolved through a universal formula.

An algorithm may classify symptoms, risks and needs consistently, but, it cannot experience the moral weight of leaving a frail person without adequate care, nor bear the consequences of being wrong.

This is why outsourcing judgement is also a crisis of political legitimacy. Public decisions are legitimate not merely because they are efficient or statistically consistent, but because they can be explained, challenged and attributed to responsible human agents. The OECD’s Digital Government Outlook 2026 similarly emphasises that managing AI's risks is integral to advancing trustworthy and resilient digital transformation. [https://www.oecd.org/en/publications/2026/06/digital-government-outlook_4585678e/full-report/adopting-and-governing-ai-in-government_7ef312a9.html]

The question, therefore, is not whether AI should assist us. It should. The question is what must never be surrendered.

We become intellectually dependent on AI when human oversight exists only in name; when professionals cannot challenge its recommendations, citizens cannot understand or appeal its decisions, and no one takes clear responsibility for the outcome.

We may outsource calculation, sorting and routine analysis; but judgement must remain an active human practice, or else, we risk building institutions in which people still make decisions in name, while machines increasingly make them in substance.