"garbage in, garbage out." — Joseph Stiglitz
Joseph Stiglitz and Maxim Ventura-Bolet’s September 2025 economic model
In a September 2025 report, Nobel Prize-winning economist Joseph Stiglitz and Columbia University colleague Maxim Ventura-Bolet modelled the information ecosystem as an economic market and reached a stark conclusion: without government intervention, markets will systematically produce more disinformation and less truth, and artificial intelligence will compound that process. Their work links everyday impressions about an unhealthy information environment to concrete incentives driving producers, platforms and consumers.
The Australian public’s view — an ANU survey of 20,000
That intuitive alarm has empirical ballast. An Australian National University survey of 20,000 average Australians found disinformation consistently ranked among top national security concerns; participants rated it as more serious than the prospect of a foreign military attack. Respondents described feeling overwhelmed by volume, unable to distinguish truth from falsehood, and manipulated by algorithms — a lived experience that Stiglitz and Ventura-Bolet set out to explain with economic theory.
How digital platforms reshaped incentives
The economists trace the decline to a shift in the market structure for information. For much of the 20th century, advertising revenue kept quality journalism — expensive to produce — financially viable. The arrival of digital platforms created a new middle layer between newsmakers and readers. Platforms aggregated and made content searchable, free for consumers, but transformed the business model: platform owners make money from engagement through ads and data collection, and every second on a platform generates revenue for the platform while a visit to the original news site generates revenue for someone else.
Stiglitz and Ventura-Bolet show platforms discovered that outrage and emotion keep users engaged longer than calm, accurate reporting. Algorithms optimised for attention therefore reward provocative content — whether true or not — while nuanced, verified sources and public-interest journalism become relatively unprofitable. As users adopt AI overviews or social feeds as their habit, original content producers lose visits and revenue; Google search and Facebook feeds became that habit before many noticed the cost.
How AI multiplies the market failure
The report argues AI both exploits and accelerates existing incentives. AI systems can produce legitimate content by scraping journalism and summarising it without paying the original producer, and they can also produce convincing fake content cheaply and quickly. Stiglitz notes AI is optimised for efficiency rather than truth: it is “not interested in whether content is truthful or untruthful.” Because AI is limited by the quality of its training data, feeding it distorted or unreliable information leads it to produce distorted content — hence the adage “garbage in, garbage out.”
The model shows damage compounds as cheap, infinitely scalable AI-generated disinformation floods the ecosystem, reducing incentives for producers to invest in costly, high-quality reporting. It also models polarisation: once disinformation gains ground, audiences seek confirming content, creating demand for more disinformation, which in turn earns revenue for its producers and drives quality news organisations further out of business.
Policy prescriptions the economists propose
Stiglitz and Ventura-Bolet argue the market will not self-correct because no actor is currently incentivised to change course. They conclude that individual choices — for instance, people opting out of social media — will not alter platform or AI company incentives. The solution, they say, is government intervention: regulated platform accountability for content amplification, enforced obligations to address coordinated disinformation campaigns, and intellectual-property protection for news producers.
What this means for quality news organisations, platforms, and the public
- Quality news and public-interest journalism: The model predicts shrinking revenue and greater pressure to curtail expensive, verified reporting unless intellectual-property protections and funding mechanisms are enforced.
- Platforms and AI companies: Without regulation, platforms and misinformation producers stand to benefit economically from attention-driven engagement and cheap AI content; voluntary measures or memoranda of understanding are insufficient to change incentive structures, the economists argue.
- The public: Australians already report feeling overwhelmed and manipulated; the model warns that without regulatory change the information landscape will become more polluted, fragmented and manipulative as AI-generated low-quality content proliferates.
Australia has already taken steps: the report notes that Australia "has a good record of leading global efforts to make digital platforms pay for news content," and that in April the government signed a memorandum of understanding with Anthropic, signalling that at least one frontier AI company is willing to engage on responsible development. But Stiglitz and Ventura-Bolet caution that a memorandum is not regulation and goodwill is not an incentive structure. The challenge, they write, is to lock in these efforts before the damage becomes irreversible.
Stiglitz and Ventura-Bolet convert a common unease into a policy problem framed by incentives: unless the rules of the information market change, the economics will continue to reward cheaper, attention-grabbing falsehoods and punish costly truth-telling. The choice Australia faces is concrete — to translate engagement into accountability, or to watch the market finish what it has begun.
