Australian politicians lead the world in speeches assessed to be written by AI, according to a recent Economist article. Politicians themselves have decided their speech-writing skills, or that of their staff, can be replaced with AI — so what other parts of their jobs could, or perhaps should, be done by machines? Perhaps the entire role of politician could be handed to agents developed from large language models.
After all, politics itself is a form of large language model. Politicians train on a vast body of political speech to learn the basic structures of political and policy discourse, with outputs — in the case of the major parties — carefully filtered to comply with those structures. Politicians respond to prompts, albeit usually far more poorly than chatbots. They share with AI a frustrating need to tell audiences what they think they want them to hear. They hallucinate facts and enthusiastically share them, pack ideological biases reflecting their training, and echo each other in circular discussions that close off the possibility of innovative thought.
The idea isn’t especially new. Albania has an AI “Minister of State for Artificial Intelligence” designed to curb corruption in public procurement — albeit with apparently slow progress. But breaking down the roles of politicians can take us some way into a serious consideration of what we actually need humans to do.
Related Article Block Placeholder
Article ID: 1267904
All politicians have several functions. One of them is representational: to be a two-way conduit between their electorates and the machinery of government, representing the views of their electors within the governmental apparatus, and providing a point of contact for their electors seeking access to that apparatus. Much of an MP’s time is taken up helping constituents looking for some form of government service or assistance. They also have a representational role within their electorates, representing their party or the parliament itself to the public.
But the representational role varies greatly. Senators notionally represent states, but have far less of a role helping constituents. In the Australian context, compared to, say, the US Senate, state representation is notional indeed. Major party senators are far more likely to adhere to their own party’s policies than aggressively represent their home state’s interests.
Similarly, major party MPs are far less likely to represent the views of their electorates than adhere to their parties’ platforms. Instead, they provide voters with an avatar for their party’s policies, identity and leadership to sign up to or reject. In contrast, independent MPs, and especially community independent MPs, tend to base their politics on representing their electorates much more fully than major party MPs, and have far greater capacity to respond to issues in line with their constituents’ views rather than the diktats of their party.
MPs and senators also have parliamentary roles: lawmaking — which the majority of modern major party MPs have virtually nothing to do with other than voting how they’re told — and accountability. There are parliamentary committees to operate, with Senate committees — and especially references committees not controlled by members of the governing party — having the most relevance to democratic accountability, scrutiny of the executive and better governance. All non-government MPs and senators play a role in accountability, and crossbench senators, and occasionally opposition senators, may play a role in improving legislation.
Then there is the political role: forming and maintaining governments and oppositions, which outside periods of minority government is only for major party MPs, determining party leaderships and promoting internal groupings and alliances, and promoting oneself.
A quick assessment of these four roles suggests AI could plausibly replace major party politicians, but not independents. A chatbot could undertake representational duties of the kind a major party MP offers their electorates, providing faster and perhaps more accurate advice to constituents on what government assistance is available for them. An AI local representative could fire off a letter on behalf of a constituent to a minister or department instantaneously. A chatbot could also handle the other end of the representational role, synthesising constituents’ views on specific issues.
Related Article Block Placeholder
Article ID: 1267539
For a major party MP, however, those views are less relevant than the party line. Only an independent MP provides a genuinely representational role for their electors within parliament and exercises real judgment about where to land on a complex issue, one that AI would struggle to mimic.
The legislative function would be easily replicated for major party MPs, especially Labor MPs, who simply vote as directed by the party leadership. But AI would struggle to accurately replace MPs in parties like the Coalition, which allows crossing the floor, and especially independent and minor party MPs more likely to assess each piece of legislation on its merits.
Similarly, non-government MPs and senators’ accountability activities in parliamentary committees would be difficult for AI in its current form to provide: the process of interrogating officials with an agenda to hide embarrassing information requires not merely the capacity to synthesise large volumes of information but also a creative appraisal of how language is being used to obscure information.
Politicians of a party in government, however, could easily be replaced by AI. A chatbot with orders to interject on non-government senators during estimates hearings, and use up precious Senate estimates time by asking Dorothy Dixers of ministers and officials, would be reasonably straightforward to create.
The political role would be harder for AI to provide, even for major party MPs: it could follow instructions from the party leadership, but working out when and how to replace that leadership would be more complex. As party factions are now as based on personality as they are ideology, especially within Labor, even an AI agent asked to be, say, a Socialist Left MP might struggle to function within a complex factional and sub-factional world. And while it could issue a stream of media releases and perform interviews, a chatbot would likely struggle to effectively promote itself.
Related Article Block Placeholder
Article ID: 1267451
But that political role is the least necessary of those functions. The factional activities of an MP don’t serve the public, nor do their efforts at self-promotion. An AI major party MP that didn’t undertake such functions would be no loss to electors. But it could deliver much the same as human MPs — reliably vote for legislation, stick to the party line regardless of the wishes of its constituents, rigidly adhere to government talking points, issue banal media releases, and direct constituents to government services.
AI agents would be particularly effective as major party senators, given their limited constituent role. AI flaws such as hallucinating facts and inbuilt bias would provide an agreeable mimicry of the existing human model. For a political party as homogenised and dissent-free as the Albanese-era Labor party, most MPs may as well be chatbots.
Would the voters of the diminishing number of electorates held by Labor or the Coalition actually notice if their local MP was an AI agent rather than a human being? Would they receive poorer service or representation than they do now? Or would AI provide a far cheaper way to deliver the same outcomes?
At each election, the major parties could construct an AI agent programmed to stick to the party line (perhaps with a factional dimension added, as agreed by the party leadership), give it a name and avatar and submit it to the voters under suitably amended electoral laws. It could then compete against independent human candidates offering a more genuine representational role to voters.
Part II up next: Ministers and staffers — the anti-corruption case for AI
