The Fifth Check, Part Two: What Would It Actually Take?

Australia’s Constitution was built on an assumption that has quietly stopped being true: that citizens can see clearly enough what government is doing to hold it to account.

Part One of this series documented how the information environment that assumption depends on, a functioning press, independent research, an alert public, has been eroded by structural forces that no single reform will undo. This piece asks the harder question that follows: if the practical conditions for democratic accountability have degraded, is there any mechanism that could partially restore them, and could such a mechanism survive the very dynamics that degraded them in the first place?

That is not a rhetorical question. It has a partial, conditional, carefully qualified answer. What follows is an attempt to work through it honestly, offered in the conviction that honest inquiry is itself a form of service, and with genuine openness to correction from any reader who finds the reasoning wanting.

Key points

• Australia already operates referendum machinery for constitutional amendments. The question is whether to extend and strengthen it, not whether to import something foreign.

  
  

• The 2023 Voice referendum is examined here not as a verdict on the Voice, but as the clearest recent case study of what happens when the institutional ecosystem surrounding a public vote is not designed to guarantee every citizen a neutral, both-sides information baseline.

  
  

• The ‘money and media will dominate it anyway’ objection is real. But it relocates the problem into a set of specific, answerable design questions, and that relocation is itself progress, because it makes the capture dynamics visible and therefore contestable at the level of institutional architecture rather than ambient narrative.

  
  

• Seven design choices would need to hold for any citizen-initiated mechanism to function as a genuine fifth check. Most commentary on referendum reform discusses one or two. This piece addresses all seven, including the one almost nobody talks about: the agenda-setting power exercised before any formal legislation is passed.

  
  

• The Christian account of institutional fallenness is not a reason to despair about these mechanisms. It is the reason they need to be built with clear eyes, watched constantly, and never mistaken for self-executing safeguards, a conviction this series has carried from its first article to this one.

1. Where Part One Left Us

Part One closed with a question rather than a verdict: if the practical conditions for what political theorists call the ‘fourth check’ on government, an informed public capable of perceiving and judging what their government does, have quietly eroded, what, if anything, could function as a fifth check in their place?

I ruled out three answers explicitly: state control of the information environment, the abolition of media companies and think tanks, and the fantasy that citizens could somehow audit everything themselves. None of those are this series’ position. Readers new to the series may want to begin with Part One, and with the theological and constitutional groundwork laid in Christ at the Centre.

What I proposed looking at instead was narrower: mechanisms that place foundational questions back before the public directly, rather than leaving them entirely in the hands of the same institutions Part One described. Not whether any such mechanism is perfect, it isn’t, or whether it eliminates the dynamics Part One described, it doesn’t, but rather whether a well-designed direct mechanism could be meaningfully more resistant to those dynamics than the representative alternatives we currently rely on, and what ‘well-designed’ would actually have to mean for that to be true.

Citizen-initiated referenda, or CIR, the idea that ordinary citizens can collect enough signatures to force a public vote on a law or constitutional question, could extend a principle already present in section 128 of the Commonwealth Constitution (the section of our Constitution that already requires major changes to be put to the people). CIR is named as one candidate among possibly several. This piece takes that candidate seriously enough to stress-test it.

One further thing worth stating plainly: the question of whether CIR could function as a fifth check and the question of whether the fourth check could simply be repaired are not mutually exclusive, but they are different questions. Strengthening media diversity, broadening Freedom of Information reform (which gives citizens and journalists the right to access government documents), improving civic education, tightening disclosure requirements, all of these are worth pursuing and none of them are this series’ opponents.

I examine CIR not because it is the only possible response to Part One’s diagnosis, but because it represents a different family of tool: one that changes the very shape of how a decision gets made, rather than merely the quality of the information flowing into it.

2. The Swiss Model, Briefly

Switzerland runs the most developed system of direct democracy among established democracies, and it is worth describing precisely rather than impressionistically, because the precision matters for what follows.

Swiss citizens have three main tools. The optional referendum allows 50,000 citizens to collect signatures within 100 days of a law being passed and force a public vote on whether to keep it. The popular initiative allows 100,000 citizens to put a proposed change to the Constitution directly to a nationwide vote, bypassing Parliament. Some matters, including all constitutional amendments, go automatically to a public vote without anyone needing to collect signatures. Put together, Swiss voters are asked to decide on national questions up to four times a year.

Two design features matter more than the voting mechanism itself. First, the Federal Chancellery (Switzerland’s equivalent of the Australian Electoral Commission) mails an official explanatory booklet to every household before each vote. It contains both sides of the argument in plain language, alongside the actual text of what is being voted on. It has run since 1977. Independent research has found the booklet's content to be the greatest single influence on voters in the political centre, with effects across all political camps. It is not a box-ticking formality.

Second, Switzerland introduced campaign finance transparency rules in October 2022, requiring public disclosure once spending on a vote exceeds around CHF 50,000 (roughly AU$85,000), and naming any individual donation above CHF 15,000 (around AU$25,000). From 2024 this extends specifically to popular votes, not just elections.

Together these two features are the closest thing on offer to a documented answer to the exact question this piece needs to ask: what stops a direct vote from simply becoming the next thing a concentrated media and funding environment captures.

3. The Obvious Objection

If the information environment is already concentrated and capturable, as Part One documented at length, why would a referendum campaign be any different? Won’t it simply be won by whoever has the most money, the loudest media presence, or the most favoured messaging online, the same closed loop Part One described, just relocated to a single vote?

This is a fair objection, and any version of this argument that doesn’t take it seriously isn’t worth making.

The weak response is that a direct vote is inherently more resistant to manipulation because it removes middlemen. A concentrated media and funding environment can shape a single yes/no question just as easily as it can shape an election, possibly more easily, since a referendum concentrates attention and money into a short, high-stakes window. Section 4 documents this using two Australian cases.

The honest response is that CIR doesn’t solve this problem by existing. It only reduces the public’s reliance on manipulable sources of information to the extent the surrounding mechanism is deliberately designed to do so. What a well-designed CIR process can do is ensure that every household has access to a guaranteed baseline of information, plain-language, both-sides, government-administered, that no media outlet, campaign donor, or social media platform can entirely prevent them from receiving.

That is a more modest and more defensible claim than ‘CIR neutralises media influence.’ It raises the floor. Whether raising the floor is enough is the question Sections 4 and 5 are trying to answer.

A third observation worth naming, held lightly: when citizens retain a direct democratic trigger, it changes the stakes of the information environment between votes. The cost of being caught distorting a question the public can directly adjudicate is higher than the cost of shaping background public opinion over years with no such mechanism in place.

This is untested and should not be overstated. But it is worth naming as a direction of argument.

4. Australia Already Has a Version of This, And Its Weaknesses Are Documented

Section 128 of the Constitution requires that any change to the Constitution be approved by the Australian people in a referendum. The Referendum (Machinery Provisions) Act 1984, the federal law governing how referendums are actually run, requires the Australian Electoral Commission to distribute an official Yes/No pamphlet to every household.

For the 2023 referendum, that meant close to 13 million copies printed and mailed. This is Australia’s own version of the Swiss explanatory booklet, government-mandated, both-sides, mailed to every household. Not a hypothetical import. A piece of machinery Australians already trust enough to use for the most consequential votes in the country’s constitutional life.

It is a guaranteed platform, not a guaranteed accuracy mechanism. The AEC has stated plainly that it ‘had absolutely no lawful authority to change or “fact check” the material’.

“The official Yes and No cases were written and authorised by parliamentarians. The AEC had absolutely no lawful authority to change or ‘fact check’ the material regardless of how strongly some people may feel about it.”

Both sides get space. Neither is required to be correct. Two Australian cases show what that gap costs in practice.

The 2013 case: the crude loophole.

The Act formally prohibits the federal government from funding Yes or No campaigns.

But in 2013, the Gillard government announced an allocation of $10 million to the Yes campaign and $500,000 to the No campaign as a ‘public information’ allocation, a twenty-to-one disparity that seven Coalition senators correctly identified as bad faith, and which contributed to the referendum bill’s collapse in the Senate.

Whether the funds were fully disbursed before the referendum was abandoned is not clearly documented, but the announced allocation alone illustrates the gap: what the rules prohibit and what ‘information spending’ can cover are far enough apart to accommodate a twenty-to-one disparity without technically breaching a single provision.

The 2023 case: the sophisticated version, an information-environment case study.

What follows is not an assessment of the Voice proposal itself, and the people named in it are not presented as villains. Each of them, as this series has always insisted, bears the image of God, and the account below is of institutional dynamics, not of individual moral failure. It is an examination of what happens to the information architecture of a public vote when no rule constrains the institutional ecosystem from deploying for one side. The 2023 referendum is the most fully documented recent example.

The Albanese government initially proposed eliminating the Yes/No pamphlet entirely, moving to remove the floor of guaranteed information access from a referendum it was simultaneously campaigning to win. The bill would have suspended section 11 of the Act, the specific provision requiring production and distribution of the official Yes and No case to every household, and would have removed the prohibition on the government spending public money to promote its preferred outcome. The pamphlet was reinstated under pressure. But the attempt itself documents a specific institutional incentive to weaken the neutral information mechanism precisely when it matters most.

The Prime Minister spoke at the Yes23 campaign launch and toured the country in the weeks before the vote. The Australian Public Service Commission issued formal guidance to federal departments framing the No campaign’s arguments as a source of psychological distress for government employees requiring managerial sensitivity and flexible leave, government machinery formally treating one side of a democratic question as a welfare risk rather than a legitimate position.

The philanthropic ecosystem that donated the majority of the Yes campaign body's approximately $47 million is not as separate from the government's policy networks as the private/public funding distinction implies. The Paul Ramsay Foundation's own public statement confirms it donated $7 million to the Yes23 campaign body, representing around 4% of its total approved giving for 2023, a year in which it committed $52.2 million to First Nations-led organisations.

The Australian Communities Foundation ran its own dedicated grants pool channelling funds directly to Yes23 campaign facilitators and affiliated community organisations. Full donation figures are recorded in the AEC's referendum financial disclosure register. These are organisations that receive government grants, sit on government advisory bodies, and depend on government goodwill for program continuation.

The media: institutionally polarised, not uniformly Yes.

The ABC’s own internal review found that 51% of the voices featured in its coverage were labelled Yes versus 23% No, measured by share of voice across broadcast and digital platforms (content monitoring by Isentia from 30 August). 79% of the 383 complaints to the ABC Ombudsman cited a favoured Yes position.

University of Adelaide research into News Corp's referendum coverage found that 70 per cent of commentary words argued No, with commentators presenting No arguments in 88 per cent of analysed content.

The Nine/Fairfax mastheads (The Age, Sydney Morning Herald) ran modestly Yes at 53/47.

The ABC’s coverage aligned more closely with the government’s preferred position than with its opponents, at a time when the government that funds it was actively campaigning for Yes. Across the media landscape as a whole, no major outlet was primarily interested in helping ordinary Australians think clearly about a significant constitutional question. Each was running its institutional position. Australians caught between them had no guaranteed neutral floor of information, because the government had recently attempted to remove the mechanism that would have provided it.

The evidence suggests this pattern would appear under any government pursuing any cause with sufficient institutional weight behind it. It is structural rather than partisan.

What both cases establish.

Together they invite the question of whether the design choices in Section 5 are responses the mechanism cannot do without, rather than optional refinements. Neither case is captured by asking simply whether direct campaign funding was provided.

One further piece of context: only 8 of Australia’s 45 constitutional referendums have ever passed, and the historical record suggests that having both major parties support a proposal has been more decisive than campaign spending in determining the outcome.

Despite the full institutional ecosystem behind Yes in 2023, 60 per cent voted No, suggesting the mechanism retains a residual resistance to institutional capture that is worth preserving and strengthening.

5. If Australia Were to Trial a Fifth Check, Design Choices That Would Need to Hold

Rather than a single finished proposal, this section sets out the specific, separable design choices required for CIR or its more modest cousin CIV (explained below) to function as a genuine fifth check rather than a weaker replay of the capture dynamics Parts One and Two have documented.

1. Start with a veto mechanism, then extend.

Full CIR, which gives citizens the power to initiate laws or constitutional changes that Parliament has not passed, would almost certainly require changing the Constitution itself, using the existing section 128 referendum process. A more achievable starting point is a Citizens Initiated Veto, or CIV: giving citizens the power to repeal a law Parliament has already passed, by collecting enough signatures to trigger a binding public vote on whether to keep it. This would follow the Swiss optional referendum model and could potentially be introduced by ordinary legislation, without requiring a constitutional change.

Australians for Better Government have recently called for precisely this. ABG President Steven Tripp describes it as a mechanism to ‘repair faith in democracy’ by ensuring significant legislative changes ‘cannot be imposed without genuine popular consent.’ A CIV is the gateway: it tests the machinery, builds public confidence, and demonstrates appetite before any broader CIR requiring a constitutional referendum is attempted.

Even this more modest scheme would face political and drafting hurdles, and its first activation would itself test whether the public is ready to use it as designed rather than as a partisan instrument. A legislative CIV is the starting point rather than the destination, and without the constitutional entrenchment that Point 7 addresses, it remains vulnerable to the same parliamentary override it is designed to check.

2. Prohibit government institutional deployment for either side during a referendum campaign.

This is the design gap the 2023 case exposed most clearly, and it goes beyond the existing prohibition on the government directly funding Yes or No campaigns. A comprehensive rule would need to cover: ministers using their official platforms and publicly funded travel to actively campaign for one side; government agencies issuing guidance that frames one side of a democratic question as a welfare or safety risk; and philanthropic organisations that are structurally dependent on government grants and goodwill using those resources for referendum campaigns.

None of these are currently covered by any rule. All of them functioned as campaign resources in practice in 2023. This rule would require careful statutory definition and an independent enforcement body with real authority, because without both, well-resourced actors will find a way around it just as they found their way around the direct-funding prohibition.

3. Add a neutral-information requirement to the pamphlet, not just a both-sides requirement.

The gap between ‘both sides get space’ and ‘an independent body must also provide a plain-language, fact-checked explanation of what the proposal would actually do in law and in practice’ is material. The AEC, or an equivalent independent body, would publish that neutral explanation alongside, not instead of, the partisan Yes and No cases. A third voice: the one whose only job is accuracy rather than persuasion. This mechanism should be constitutionally protected from the kind of attempted removal that occurred in 2022. It is the floor the whole argument depends on.

4. Close the gap between prohibited campaign funding and permitted ‘information’ spending.

Following Switzerland’s 2022 reform: public disclosure once campaign spending exceeds a threshold, and disclosure by name of any individual donation above a set amount, including from philanthropic foundations and the donor-advised funds they use to pool and distribute contributions.

In the Australian context: clear spending caps on what the government can call ‘public information’ spending around a referendum, mandatory disclosure of all campaign expenditure above a threshold from any source, and explicit funding parity rules where any government money touches the campaign environment. Both the 2013 crude loophole and the 2023 sophisticated philanthropic ecosystem need to be designed against. Closing only one leaves the other available.

5. Pair CIR or CIV with deliberative supplements.

Citizens’ assemblies, where a representative, randomly selected group of ordinary people deliberates on a question with expert input before it goes to a full public vote, provide a tested public benchmark for what an informed judgment looks like. Ireland used deliberative bodies ahead of both its 2015 marriage equality referendum, where the Convention on the Constitution (2013–2014) considered and recommended the change, and its 2018 abortion referendum, where the Citizens' Assembly (2016–2018) deliberated on the Eighth Amendment.

Australia has its own version of this kind of body: the 1998 Constitutional Convention, where 76 of 152 delegates were elected by popular postal ballot, is the closest Australian precedent for a citizen-deliberative body with formal constitutional standing. The legitimate concern about both mechanisms is that they get framed as ‘unrepresentative elites making decisions for the rest of us.’

The design response: a transparent selection process open to public scrutiny, published deliberations, strict limits on what the body can decide, and a clear statement that it is there to inform the vote, not replace it. Its output sits on the public record as a mandatory addendum before the campaign begins, not as a binding decision.

6. Define what counts as ‘legislation’, including non-binding resolutions and motions.

A South Australian Legislative Council motion examined elsewhere in this series shows the gap: the chamber voted in principle to explore a regulatory framework for digital platforms, creating no actual law and binding no future government to anything specific. But by the time any formal bill implementing that direction arrives, opposing it looks like re-litigating a question the chamber already settled, even though no threshold, pamphlet, or disclosure rule proposed above would have applied to the motion that did the actual settling.

When a chamber signals a direction through a non-binding vote, the later formal legislative stage can become a fait accompli, a done deal, rather than a genuine open decision point. That makes the question of what triggers a fifth check mechanism not merely a technical drafting issue but central to whether the mechanism can reach the actual site of agenda control, which is the step before the legislation, not the legislation itself.

A CIV as currently proposed would not reach this stage, which is precisely the point: any fifth check mechanism that defines its trigger narrowly enough to exclude where the real decision was already made is not yet a full fifth check.

7. Protect CIR or CIV results from immediate institutional retaliation.

If citizens successfully pass an initiative or veto, what prevents the High Court (Australia’s highest court, which decides whether laws are constitutionally valid) from striking it down, or a hostile Parliament from quietly amending or repealing it six months later by an ordinary vote?

A fifth check that can be immediately dismantled by the institutions it is designed to check is not a structural safeguard, it is a temporary political event. A serious design would need to address at minimum: (a) whether citizen-initiated outcomes receive some form of enhanced protection requiring a higher parliamentary threshold to reverse (for example, a two-thirds majority rather than a simple one); (b) what standards courts should apply when reviewing citizen-initiated results; and (c) whether a defined period should apply during which Parliament cannot amend or repeal the result without a fresh public vote.

Some of these protections would themselves require constitutional change, which is not a reason to abandon them, but a reason to be clear-eyed about the sequence. History gives little reason to expect the institutions being checked will simply accept the check.

These seven choices are not mutually exclusive. Points 2, 3, and 4 form an interlocking defence that probably needs to be adopted together to be meaningful: rules on institutional deployment, a neutral information requirement, and funding disclosure each leave a gap the others fill. Point 6 is a precondition for all of them: none of it reaches the actual site of agenda control if ‘legislation’ is defined narrowly enough to exclude the step where the real decision was already made. Point 7 is the test of whether any of the above is durable.

One open question that any federal proposal will eventually need to confront: Australia’s Constitution currently requires that referendum questions be approved not just by a national majority of voters, but also by a majority of voters in at least four of the six states.

This double-majority requirement is why even popular referendum questions can fail if they don’t carry enough states. Whether to apply some version of this to CIR or CIV questions, and to which types of questions, is a live design issue I do not resolve here, but readers may want to consider their own instincts.

6. What This Doesn’t Solve

A former director of the Swiss Federal Audit Office has publicly stated that the 2022 transparency rules contain loopholes and that the results of spot-checks on disclosure compliance are not published.

Even the most developed version of this mechanism in the world has acknowledged, unresolved gaps. That is not a reason to abandon the idea, it is proof of the point Section 8 makes: no mechanism is self-executing. A system with disclosure loopholes is still more resistant to invisible capture than a system with no disclosure requirement at all. But anyone proposing this mechanism honestly has to concede it would need ongoing scrutiny of its own.

Part One described a daily problem, the ambient, background erosion of the information conditions that constitutional accountability depends on. CIR and CIV are episodic. They create structured moments of officially resourced, both-sides deliberation.

They do not improve the information environment on the ordinary days between votes. What they may do, if well designed, is create a kind of anchor, a guaranteed channel that exists outside the normal media and funding ecosystem, and whose very existence changes the stakes of that ecosystem even when it is not being used. A structural floor, not a ceiling, against which the ambient dynamics must now work harder.

7. Two Other Families of Tool, And Why This Piece Focuses on CIR

Recall elections.

Nineteen US states allow voters to remove a sitting governor from office before their term ends, by collecting enough signatures to trigger a special election. Eight of those states require specific grounds such as misconduct; the other eleven allow any registered voter to initiate the process for any reason.

Recall provisions answer a different question from CIR. CIR asks: ‘should this law exist?’ Recall asks: ‘should this person continue to hold office?’ Part One documented a structural problem: the erosion of conditions for perceiving and judging government action across multiple institutions over a long period. No recall of any individual officeholder addresses that. A mechanism that places foundational questions directly before the public, with guaranteed information architecture, at least partially does. That is why CIR is the candidate I stress-test here.

The South Australian entrenchment finding.

While fact-checking the SA Voice section in Part One, a finding surfaced that sharpens the ‘forms that exist but don’t do the work’ pattern running through this piece. The Constitution Act 1934 (SA) locks in, via a referendum requirement, the existence and term of both Houses of the SA Parliament and the rules for redistributing electoral boundaries. What it does not lock in is citizens’ substantive rights or their standing in relation to those institutions, which is why the Voice Act 2023 could proceed by ordinary parliamentary majority with no referendum required and none possible to demand (the SA Constitution, unlike the Commonwealth's, does not require a referendum to be amended in this area)."

The NSW precedent, confirmed by the High Court in 1931 and on appeal to the Privy Council in 1932 in Trethowan's case, established that state parliaments could choose to lock certain things in place by requiring a referendum before they could be changed, and that this lock was legally enforceable. The pattern across SA, NSW, and Queensland shows how that power has been used: when Australian state parliaments have reached for the referendum lock, the one mechanism that is supposed to hand a decision to the people rather than leave it with politicians, they have used it to protect Parliament's own continuity and architecture. Not citizens' rights. Not representation. Not recognition.

The institutions, entrenching themselves, using the one tool that exists to let the people object. That is not conspiracy. It is the entirely predictable outcome of institutions designing their own constitutional protections. Which is precisely why Point 7 in Section 5 matters: a fifth check mechanism that can be immediately unwound by the institutions it is designed to check risks following exactly the same pattern.

8. What Christianity Brings to This Argument

Everything in the preceding sections has been argued on structural and historical grounds, and deliberately so. The argument does not require a Christian starting point to hold. But for this series, and for me personally, there is a deeper grounding that it would be dishonest to omit. Not as a proof. Not as a conclusion only Christians can reach. But as the reason I find myself returning to these questions with something more than civic concern. The fuller theological frame is in Christ at the Centre, and in the series that preceded this one, beginning with What Is a Woman?.

There is a thread that runs through human moral reasoning across cultures and centuries that recognises, without needing the Christian vocabulary to name it, that power tends to corrupt those who hold it, that institutions tend to serve themselves before they serve those they were built for, and that the people most vulnerable to concentrated power are rarely the people designing the systems that exercise it.

The Stoics named it. The Roman legal tradition encoded protections against it. Locke built a political philosophy on it. The framers of the American Constitution made it the central premise of everything they designed: that no human being and no human institution can be trusted with unchecked power, because the tendency to abuse it is not an aberration but a feature of the human condition.

What the Christian tradition adds is not a different conclusion but a more specific account of why it is true. The word theologians use is fallenness, not a counsel of despair, but a clear-eyed description of the gap between what human beings were made to be and what, left to their own devices, they reliably become. Every institution ever built, Parliament, the press, the courts, the church, the think tank, the regulator, the referendum campaign itself, carries that fallenness into its design. It will tend, over time and under pressure, toward self-preservation rather than service, toward opacity rather than transparency.

Madison did not give America three branches of government because he trusted the people who would fill them. He did it because he didn’t, because he understood that the only reliable check on fallen human nature is another fallen human institution with sufficient power and sufficient incentive to push back. Ambition counteracting ambition. Not because anyone will be virtuous, but because the structure makes virtue less necessary and vice less rewarding. This is not cynicism. It is the premise of every serious constitutional design in history, from Magna Carta (the 1215 charter that first established that even a king could be bound by law) to the separation of powers we take for granted today.

The argument of this series is that the structure Madison relied on, an informed public as the ultimate check on all the others, has itself become vulnerable to the same tendency it was designed to guard against. That is not a claim about conspiracy. It is a claim about fallenness operating at institutional scale, entirely predictably, in exactly the way the Christian tradition has always said it would.

The 2023 Voice referendum is a case in point, and what it illustrates is structural, not partisan. No single actor needed to conspire. The Prime Minister campaigned because he believed in the cause. The ABC covered it according to its institutional culture. The philanthropic foundations gave because they shared the government’s policy direction. The public service issued wellbeing guidance because it was genuinely concerned for its First Nations staff.

Every one of them did what institutions do: acted according to their own values, interests, and incentive structures within the systems they inhabited. The same cumulative pattern would appear under any government, of any political persuasion, pursuing any cause with sufficient institutional weight. That is fallenness at scale. It doesn’t require malice. It only requires ordinary people doing what comes naturally within fallen structures.

Which means the response cannot be to trust that any institution, including CIR or CIV, will behave better than its incentives predict. The Swiss transparency rules have loopholes. The Australian Yes/No pamphlet was nearly abolished. Any mechanism built would carry its own vulnerabilities the moment it became worth capturing. This is not a reason to abandon the proposal.

It is the reason Section 6 exists: to name those limits honestly, and to insist that no mechanism is self-executing. Every safeguard requires a public capable of watching it. And that requires something constitutional design alone cannot manufacture, the kind of civic character that knows its own tendency toward comfort and convenience, and chooses the harder work of paying attention anyway.

For readers who share the Christian framework, that civic character has a name and a source: the image of God in which every human being is made, what theologians call the imago Dei, the conviction that every person, not just the powerful, carries the capacity to reason, to judge, to distinguish truth from manipulation, and to act on what they see. That capacity is not extinguished by fallenness. But it requires cultivation, in the church, in the family, in the civic associations Tocqueville saw as the essential connective tissue of democratic life, that actively form it rather than passively assuming it will persist.

For readers who do not share that framework, the argument lands in the same place by a different road. What theologians call general revelation, the idea that some basic moral knowledge is available to all human beings through conscience and reason, not just through Scripture, is described by Paul in Romans 2:14–15 and Acts 17:26–27. The sense that power ought to be answerable, that truth matters, that the vulnerable deserve protection, is not the exclusive property of Christians. It is the common inheritance of humanity.

What I am asking is whether Australia’s constitutional arrangements are currently making it easier or harder for that common inheritance to do its work. The evidence assembled across Parts One and Two suggests the answer is harder. And the proposal under examination is whether a well-designed direct mechanism could begin to change that.

I am not claiming CIR is a Christian solution to a constitutional problem. I am claiming that the Christian account of human nature is what makes me take both the problem and the partial solution seriously, rather than assuming that the right institutions, left to themselves, will eventually produce the right outcomes. They won’t. Not because the people in them are unusually bad, but because the people in them are entirely ordinary. Which is enough.

9. Where This Leaves Us

The argument across Parts One and Two has been narrow and deliberate. Not: Australia’s democracy is broken. Not: the media is the enemy, or government is corrupt, or citizens have been deceived. Those are the sweeping claims this series has tried hardest to avoid, because they are less provable and less useful than the more specific one.

The specific claim is this: a constitutional system can remain formally intact while losing some of the practical conditions that make it work. Part One documented that the information environment the Constitution implicitly assumes, an alert, informed public capable of perceiving and judging what government does, has been quietly eroded by structural dynamics that no single actor designed and no single reform will undo.

Part Two has asked whether any mechanism could partially compensate for that erosion. The 2023 Voice referendum is the most recent and fully documented case study of what happens when no such mechanism exists: institutional actors behave according to their interests, the information environment polarises along institutional fault lines, and the floor of guaranteed neutral information that every citizen is entitled to access nearly disappears, not through malice, but through the entirely ordinary operation of institutions in the absence of rules designed to constrain them.

The answer I reach is cautious and conditional. CIR or CIV, if well designed, with rules on institutional deployment, a neutral-information requirement, comprehensive funding disclosure, deliberative supplements, protection from immediate retaliation, and a clear definition of what decisions trigger it, could function as a genuine, if partial, fifth check. Not because it solves the problem Part One described. It doesn’t. But because it changes the architecture of certain decisions in ways that make them more resistant to the capture dynamics that architecture currently enables. It raises the floor. A higher floor, honestly acknowledged as a floor rather than a ceiling, is a meaningful improvement over what currently exists.

Australia already has the plumbing. The referendum machinery is real, and has been used for the most consequential votes in the country’s constitutional life. The question is not whether to build something from scratch. The question is whether to complete what was already started, to add the filtration to the plumbing, and to extend the principle of ‘some decisions are too important for parliament to make alone’ beyond the narrow set of questions it currently covers.

I am asking whether that is worth doing. Not declaring that it is. The difference between those two postures is, I think, the whole point of a series like this one: to sharpen the question clearly enough that readers who have never thought about referendum design, or Senate majorities, or Swiss disclosure rules, find themselves thinking about them anyway, and find themselves better equipped to ask hard questions of power, whoever holds it. Including power exercised through the mechanisms we build to hold power to account.

This is, in the end, what the series that began with What Is a Woman?, moved through Who Shapes the Child?, From Philosophy to Federal Court, Who Funds the Redefinition of Sex?, and Who Targets the Church? has always been about: helping Australians see the structures through which power moves, in whatever form those structures take, and asking whether the accountability mechanisms we have are equal to the task. The answer, in each case, has been: not quite. This piece is no different. The fifth check, if it is ever built, will also need watching.

If I have reasoned wrongly here, or missed something that matters, I am genuinely glad to hear it. That openness is not a formality. It is the same conviction that runs through every article in this series: that no writer, and no mechanism, is self-executing.

For everything that is hidden will eventually be brought into the open, and every secret will be brought to light. Mark 4:22

Until next time,

God bless,

Mark Neugebauer

———