Could Large Language Models Replace Voters, Politicians, Lawyers and Judges?

By Matthew Parish, Associate Editor

The rapid emergence of large language models has revived an old philosophical inquiry: whether political and legal authority is grounded in human judgment or whether it might one day be assumed, in whole or in part, by computational systems. Artificial intelligence is already transforming countless administrative and informational functions within states, but the more provocative question is whether she could, in time, replace the human actors who comprise a constitutional order: voters, politicians, lawyers and judges. That prospect would have seemed fanciful only a few years ago. Today it is a matter of genuine debate in academic and policy circles, although its practical, ethical and constitutional implications remain deeply contested.

There are four distinct institutional spheres to consider. Each is embedded in a long history of struggle about legitimacy, participation and accountability. Each also represents a domain where large language models possess strengths unknown to humans, yet also suffer from limitations that strike at the core of democratic and legal authority. Exploring these four domains helps us understand both the transformative potential of artificial intelligence and the limits that principled societies might sensibly impose.

Replacing Voters

Democratic systems rely upon the aggregation of individual opinions, each rooted in personal experiences, beliefs, interests and values. In principle, language models can capture an enormous breadth of social information, digest it at speeds unimaginable to a human electorate and generate a synthetic preference map for society. One could imagine a system in which voters no longer cast ballots; instead, an artificial agent infers policy preferences from an individual’s digital behaviour and expresses them in legislative form. Such a mechanism might seem attractive for its efficiency. It could also be defended on the grounds that citizens are often ill-informed or uninterested in the details of public policy.

Yet the central difficulty is legitimacy. A vote is not merely an expression of preference; it is an autonomous political act in which the voter accepts responsibility for shaping the state. If artificial intelligence were to cast votes on behalf of citizens, the participatory component of democracy would be hollowed out. The social contract would be mediated by an algorithmic oracle. Even if perfectly accurate in predicting individual preferences, such a system would obscure the moral weight of political choice. The dignity of citizenship lies partly in the acceptance of imperfect human judgment. Moreover any attempt to replace the electorate with automated proxies would concentrate power over training data, model design and parameter selection in the hands of a small technical elite. Democracy would risk becoming a machine-interpreted simulation of itself.

Replacing Politicians

Politicians (at least in democracies) serve two primary functions: they mediate competing social interests and they provide leadership. Mediation involves negotiation, persuasion and compromise; leadership involves the ability to articulate a vision, navigate crises and command public confidence. Large language models excel in the generation of coherent policy frameworks. They can simulate policy debates, test trade-offs and refine legislative priorities far more efficiently than traditional party research teams. It is therefore conceivable that future parliaments might rely heavily upon artificial advisers to design and evaluate policy proposals.

However this is distinct from replacing politicians outright. A politician’s authority derives not only from knowledge but from representation. Citizens elect individuals whose character, values and judgement they trust, even if only marginally. Leadership in moments of crisis, uncertainty or tragedy requires emotional presence, moral courage and the ability to embody collective will. A machine may generate eloquent speeches and flawless briefs, but it cannot embody the lived experiences of a community. Nor can it be held accountable in any meaningful sense. Without the possibility of removal from office, censure or resignation, political authority becomes decoupled from responsibility. Replacing politicians with artificial intelligence would therefore transform governance from a human enterprise into a technocratic mechanism. This may produce competent administration but at the cost of democratic legitimacy and human solidarity.

Replacing Lawyers

Lawyers operate within a vast terrain of interpretative practice. They identify relevant legal principles, apply them to facts, anticipate opposing arguments and communicate positions to courts and other parties. Artificial intelligence already performs some of these tasks. Document review, contract analysis and legal research can be partially automated. Indeed language models can now produce draft memos and conduct rapid comparisons across large corpora of case law.

In theory these models might become sufficiently advanced to generate complete legal arguments with precision and consistency. They would eliminate human error, manage enormous volumes of data and, in time, adapt to the evolving interpretations of statutory and case law. This would change the legal profession profoundly. Routine legal work could be done instantaneously, leaving only complex strategy and client interaction for human lawyers.

But the role of the lawyer is not purely technical. It is a fiduciary profession requiring loyalty to a client’s interests and an understanding of social, ethical and commercial context. Clients do not seek only the best possible legal argument; they seek counsel. That relationship is grounded in trust, discretion and experience. Artificial intelligence may simulate these qualities, but the simulation lacks moral agency. It cannot accept responsibility for professional negligence or ethical breaches. A system that replaced lawyers entirely would strip human clients of a personal advocate and reduce the practice of law to computational optimisation of outcomes. Whether that would constitute justice is doubtful.

Replacing Judges

The judiciary occupies the apex of legal authority. Judges resolve disputes, interpret statutes, oversee trials and articulate principles that bind society. They also exercise discretion, particularly in sentencing and the equitable adjustment of disputes. Artificial intelligence offers, at least at first glance, several advantages. A model could produce consistent reasoning, avoid personal bias, accelerate decision-making and survey vast data to identify patterns in precedent. Indeed some legal systems have already experimented with algorithmic sentencing recommendations.

The deeper issue is that judging is not only a matter of applying rules. It is a moral and civic role, requiring sensitivity to fairness, equity and the social consequences of decisions. Human judges do not merely calculate outcomes; they reason through dilemmas, grapple with ambiguity and engage with the human stakes of each case. Their authority rests on a mixture of legal skill, institutional independence and public trust. A machine cannot stand in symbolic relation to society as the arbiter of justice. It can interpret patterns, but it cannot embody conscience. Moreover the risk of hidden biases embedded in training data is significant. An automated judiciary would risk entrenching structural injustices in a way that appears objective but remains opaque and unchallengeable.

The Convergence of Automation and Authority

The question of whether artificial intelligence could replace the essential pillars of a democracy is therefore less a technical inquiry than a constitutional one. The essential weakness of artificial governance is not that models are incapable of performing the requisite tasks; rather, it is that the political and legal order is built upon principles of human responsibility. Delegating authority to a machine voids the moral contract that underpins democracy and the rule of law. Artificial intelligence can augment, advise and inform, but it cannot legitimately replace the human structures of consent and accountability.

That is not to say that large language models have no role. They may become indispensable tools for policy analysis, legislative drafting, legal research and evidential interpretation. They might help identify inconsistencies in judicial decisions or suggest avenues of legal reform. They could make democratic participation easier by providing citizens with clear, accessible information. But the final act of judgment, whether in the voting booth, the legislature, the lawyer’s office or the courtroom, must remain a human one.

The challenge for the coming decades will lie in designing institutions that harness the extraordinary analytical power of artificial intelligence without eroding the human dignity that gives political and legal authority its meaning. Replacing humanity with machinery at the summit of these institutions would produce a system efficient in operation yet impoverished in spirit. The proper goal is not substitution but symbiosis: a constitutional order in which machines serve humans, rather than humans serving the machines that they created.