When rivers degrade, pests spread or drought hits crops, nature sends a bill.

Yet it’s one rarely itemised on any balance sheet, because nature’s contribution to business remains genuinely hard to quantify.

One major obstacle is data. Businesses rarely disclose their precise operating locations, while detailed ecological information that can be linked to specific firms is scarce in most countries.

This is despite healthy ecosystems underpinning large parts of the economy, from agriculture and forestry to tourism and food production. As the US economist Herman Daly famously put it, the economy is “a wholly owned subsidiary of the environment, not the reverse”.

As part of a growing body of global research now trying to put hard numbers on what nature actually contributes to the economy, we looked at New Zealand’s case.

Our newly completed research turned up a compelling finding: firms operating in areas with richer biodiversity are measurably more productive.

Measuring nature’s value

We chose New Zealand because it publishes detailed sets of business and environmental data. That allowed us to compare company performance with local ecological conditions across different regions.

We combined measures of sales and employment with biodiversity indicators – including river health, drought risk, land use and invasive species – used as part of international reporting obligations.

We also drew on the Cobb-Douglas economic model – commonly used to estimate how labour and investment drive economic output – to help get a clearer picture of nature’s economic contribution as a factor of production.

We found businesses operating in areas with healthier ecosystems tended to generate higher sales and profits.

Across more than 117,000 observations spanning 2009 to 2022, a 1% increase in natural capital was associated with sales about 0.13% higher and profits about 0.15% higher on average. The relationship remained consistent across multiple measures of biodiversity and ecosystem health.

We also found a trade-off. Areas with more roads, buildings and commercial activity tended to have lower biodiversity scores but higher sales. In other words, businesses can still grow while degrading nature – but may lose some of the productivity benefits healthy ecosystems provide.

When green policy boosts productivity

We also tested whether major environmental policies changed this relationship.

One was New Zealand’s Predator Free 2050 programme. The other was a broader package of reforms introduced from 2017, including freshwater rules, tree-planting incentives, restrictions on offshore oil and gas exploration, limits on single-use plastics and the Zero Carbon Act.

Because these policies targeted ecosystems rather than directly subsidising firms, they helped us test whether improvements in nature were linked to changes in business performance.

We found the relationship between healthy ecosystems and business performance became even stronger following both interventions, with the productivity effect associated with 1% more natural capital increasing business performance by a further 0.05%. The effect was strongest in the year immediately afterwards.

This suggests investment in ecological restoration and protection can generate economic benefits beyond the environmental sector itself.

The strongest effects appeared in agriculture and forestry, where business outcomes are closely tied to the health of surrounding ecosystems.

Farms and forestry operations in less intensively developed areas – with lower population density and less infrastructure – showed markedly stronger productivity gains linked to natural capital.

In these primary industry regions, a 1% increase in natural capital was associated with sales that were additionally higher by 0.71% to 0.81% above the economy-wide average.

This is unsurprising. Healthy soils, clean water, fewer pests and intact native vegetation can support food and fibre production while lowering costs.

The benefits were also evident in service industries, construction and retail, although spread more evenly across a broader range of ecological factors.

An unseen benefit

These New Zealand insights are important for the growing global effort to better understand the economic value of nature. Globally, the services ecosystems provide to business are estimated to be worth trillions of dollars annually.

While new frameworks such as the international Taskforce on Nature-related Financial Disclosures are beginning to emerge, hard evidence linking ecological conditions to firm-level productivity has remained limited.

Our study suggests biodiversity is not simply an environmental concern. Differences in ecosystem health across regions and industries are associated with measurable differences in business performance.

Businesses should view the natural environment as a productive asset every bit as real as machinery or labour, not just background scenery.

And for policymakers – particularly in countries reliant on primary industries, such as New Zealand and Australia – ecological investment and economic productivity shouldn’t be taken as opposing goals.

Nature, it turns out, has been doing more economic work than some have given it credit for.

The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.

Diplomacy often masquerades as theater. And nearly nine years after his first state visit to China, Donald Trump returned to Beijing with an extended cast of characters.

Alongside the U.S. president on his May 2026 visit was a senior delegation of politicians including his secretary of defense, and a phalanx of business leaders and technology executives. It was a traveling display of American political and corporate power.

Not that the hosting Chinese were short of symbolic gestures themselves. Trump’s first China visit in 2017 had already shown how far Beijing was willing to go to turn diplomacy into theater. On that occasion, Chinese President Xi Jinping and his wife Peng Liyuan personally accompanied Donald and Melania Trump through the Forbidden City, Beijing’s former imperial palace, drinking tea inside the palace walls and taking in a Peking opera at the Belvedere of Pleasant Sounds, a Qing imperial theater built for court entertainment.

So what was being conveyed this time around? As a cultural historian of modern China, I took a peek beyond the official statements and trade headlines of the Xi-Trump summit and into the images, gestures and cultural symbolism on display.

The weight of heaven

The formal choreography began at Beijing’s Great Hall of the People, where the two leaders exchanged views on the Iran conflict, the war in Ukraine and the Korean Peninsula, among other items.

But the more interesting story of the visit, to me, was told outside the meeting room.

After their two-hour bilateral meeting, Trump and Xi paid a cultural visit to the Temple of Heaven in Southern Beijing. Built in the early 15th century, the temple is China’s most complete surviving imperial religious complex. For nearly five centuries, emperors of the Ming and Qing dynasties came here to worship Heaven and pray for good harvests.

Its most recognizable structure, the Hall of Prayer for Good Harvests, rises in three tiers of blue-glazed tiles above a marble platform, its circular form and crimson columns translating cosmology into architecture. UNESCO inscribed the site as a World Heritage Site in 1998, recognizing it as “a masterpiece of architecture and landscape design.”

When Trump and Xi posed for photographs, they were standing in a place long associated with cosmic order and the welfare of the people. To bring a foreign leader there is to invite a particular reading of the relationship: not simply as a bargain between states, but as a relationship that Beijing hopes to associate with order, abundance and peace.

There was also a more practical layer to this symbolism. The Temple of Heaven links political authority to agricultural abundance. Emperors came here to pray not for abstract harmony but for grain. That made it a pointed setting for a visit in which American agricultural exports — soybeans, grains and beef among them — were expected to matter.

For Trump, any Chinese commitment to buy more U.S. farm goods would have clear domestic political value. For Xi, the setting allowed a hard bargaining issue — farm purchases — to be translated into an older symbolic language of harvest that spoke to both domestic and international audiences.

Before Trump, Kissinger

Trump was not the first American statesman to be brought to the Temple of Heaven.

In July 1971, Henry Kissinger, then national security adviser to President Richard Nixon, arrived in Beijing on his famous secret mission — the back-channel visit that helped re-open the door between two countries that had little direct contact for more than two decades. Between tense negotiations with Chinese premier Zhou Enlai, Kissinger made time to visit the temple.

There, standing amid the old cypress groves, he was said to have been deeply moved by the timeless atmosphere of the hall and its surroundings.

The motif of old trees and deep time returned on May 15, when Xi gave Trump a rare walk through Zhongnanhai, the walled compound that now houses the core of China’s party-state leadership. Reuters reported that a hot mic captured Xi drawing Trump’s attention to the age of the trees around them — some centuries old, some said to be more than a thousand years old. When Trump asked whether Xi had taken other presidents on similar walks, Xi replied that he had only rarely.

Together, the Kissinger anecdote and the Zhongnanhai walk reveal a recurring logic in Chinese-American diplomacy: America’s fast-moving economy is invited to look at China’s sense of tradition. Xi has used this tactic with other leaders, too. When French President Emmanuel Macron visited China in 2023, he attended a guqin performance invoking the classical idea of the zhiyin — the rare listener who truly understands one’s music.

Basketball and roast duck

Trump’s visit was not staged only through imperial grandeur, however. It also moved into a more familiar register: food, sports and popular culture.

The state dinner on May 14 was another study in careful hospitality. Chefs designed the menu to honor both Chinese culinary prestige and Americans’ — and Trump’s — known preferences: Peking roast duck, crispy beef ribs, pan-fried pork bun, tiramisu and fruit and ice cream.

Trump thanked Xi for a “magnificent welcome like none other,” then replied in a language more recognizably his own. He spoke not only of power politics but of people-to-people ties: Chinese workers who helped build America’s railroads, Chinese enthusiasm for basketball and blue jeans and the sheer presence of Chinese restaurants across the U.S.

The examples were characteristically Trumpian — simple, vivid and easy to grasp. But they pointed to something important. U.S.–China relations have never been made only by presidents, diplomats and official communiques. They have also been shaped by athletes, musicians, restaurant owners, students and tourists.

The basketball reference was especially resonant. Sports have long offered a softer language for U.S.–China relations. In April 2026, just weeks before Trump’s visit, China and the U.S. marked the 55th anniversary of ping-pong diplomacy — the famous 1971 exchange in which a “little ball” helped move the “big ball” of world politics.

Basketball now plays a similar role. For many Chinese fans, the NBA is a deeply familiar world of players, teams and memories that represents the spirit of America: Michael Jordan, Kobe Bryant, LeBron James and Yao Ming. That reservoir of affection has survived even periods of political tension. Trump, in invoking it, was drawing on something real.

A second act in the US?

The main lesson of all this symbolism is that, in U.S.–China relations, atmosphere has never been secondary.

Diplomatic theater cannot settle disputes over technology or Taiwan, or determine the future of the global order. But it can shape the mood in which rivalries are managed, and the stories that leaders tell their public about what the relationship means.

And on that front, the summit worked on several levels. To the Chinese audience, it presented their leaders as confident and capable of managing a tense relationship with the U.S. on China’s own cultural terms.

For Trump and the American delegation, it offered a lesson in Chinese traditions and culture that promotes deeper understanding across political divides. And for both societies, the references for food, sports and popular culture created a more neutral ground on which connection could still be imagined.

From the 1970s opening to Trump’s 2017 visit to the Forbidden City, and from the Temple of Heaven photo-op to the walk among old trees at Zhongnanhai in 2026, cultural staging remains central to how China presents itself to America — and how America is invited to imagine China. It was announced on May 15 that Xi will pay a state visit to the U.S. in September at the invitation of Trump. If that happens, the theater of diplomacy will move to American soil, and the question will be how Washington chooses to stage China in return.

Xianda Huang does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

Diplomacy often masquerades as theater. And nearly nine years after his first state visit to China, Donald Trump returned to Beijing with an extended cast of characters.

Alongside the U.S. president on his May 2026 visit was a senior delegation of politicians including his secretary of defense, and a phalanx of business leaders and technology executives. It was a traveling display of American political and corporate power.

Not that the hosting Chinese were short of symbolic gestures themselves. Trump’s first China visit in 2017 had already shown how far Beijing was willing to go to turn diplomacy into theater. On that occasion, Chinese President Xi Jinping and his wife Peng Liyuan personally accompanied Donald and Melania Trump through the Forbidden City, Beijing’s former imperial palace, drinking tea inside the palace walls and taking in a Peking opera at the Belvedere of Pleasant Sounds, a Qing imperial theater built for court entertainment.

So what was being conveyed this time around? As a cultural historian of modern China, I took a peek beyond the official statements and trade headlines of the Xi-Trump summit and into the images, gestures and cultural symbolism on display.

The weight of heaven

The formal choreography began at Beijing’s Great Hall of the People, where the two leaders exchanged views on the Iran conflict, the war in Ukraine and the Korean Peninsula, among other items.

But the more interesting story of the visit, to me, was told outside the meeting room.

After their two-hour bilateral meeting, Trump and Xi paid a cultural visit to the Temple of Heaven in Southern Beijing. Built in the early 15th century, the temple is China’s most complete surviving imperial religious complex. For nearly five centuries, emperors of the Ming and Qing dynasties came here to worship Heaven and pray for good harvests.

Its most recognizable structure, the Hall of Prayer for Good Harvests, rises in three tiers of blue-glazed tiles above a marble platform, its circular form and crimson columns translating cosmology into architecture. UNESCO inscribed the site as a World Heritage Site in 1998, recognizing it as “a masterpiece of architecture and landscape design.”

When Trump and Xi posed for photographs, they were standing in a place long associated with cosmic order and the welfare of the people. To bring a foreign leader there is to invite a particular reading of the relationship: not simply as a bargain between states, but as a relationship that Beijing hopes to associate with order, abundance and peace.

There was also a more practical layer to this symbolism. The Temple of Heaven links political authority to agricultural abundance. Emperors came here to pray not for abstract harmony but for grain. That made it a pointed setting for a visit in which American agricultural exports — soybeans, grains and beef among them — were expected to matter.

For Trump, any Chinese commitment to buy more U.S. farm goods would have clear domestic political value. For Xi, the setting allowed a hard bargaining issue — farm purchases — to be translated into an older symbolic language of harvest that spoke to both domestic and international audiences.

Before Trump, Kissinger

Trump was not the first American statesman to be brought to the Temple of Heaven.

In July 1971, Henry Kissinger, then national security adviser to President Richard Nixon, arrived in Beijing on his famous secret mission — the back-channel visit that helped re-open the door between two countries that had little direct contact for more than two decades. Between tense negotiations with Chinese premier Zhou Enlai, Kissinger made time to visit the temple.

There, standing amid the old cypress groves, he was said to have been deeply moved by the timeless atmosphere of the hall and its surroundings.

The motif of old trees and deep time returned on May 15, when Xi gave Trump a rare walk through Zhongnanhai, the walled compound that now houses the core of China’s party-state leadership. Reuters reported that a hot mic captured Xi drawing Trump’s attention to the age of the trees around them — some centuries old, some said to be more than a thousand years old. When Trump asked whether Xi had taken other presidents on similar walks, Xi replied that he had only rarely.

Together, the Kissinger anecdote and the Zhongnanhai walk reveal a recurring logic in Chinese-American diplomacy: America’s fast-moving economy is invited to look at China’s sense of tradition. Xi has used this tactic with other leaders, too. When French President Emmanuel Macron visited China in 2023, he attended a guqin performance invoking the classical idea of the zhiyin — the rare listener who truly understands one’s music.

Basketball and roast duck

Trump’s visit was not staged only through imperial grandeur, however. It also moved into a more familiar register: food, sports and popular culture.

The state dinner on May 14 was another study in careful hospitality. Chefs designed the menu to honor both Chinese culinary prestige and Americans’ — and Trump’s — known preferences: Peking roast duck, crispy beef ribs, pan-fried pork bun, tiramisu and fruit and ice cream.

Trump thanked Xi for a “magnificent welcome like none other,” then replied in a language more recognizably his own. He spoke not only of power politics but of people-to-people ties: Chinese workers who helped build America’s railroads, Chinese enthusiasm for basketball and blue jeans and the sheer presence of Chinese restaurants across the U.S.

The examples were characteristically Trumpian — simple, vivid and easy to grasp. But they pointed to something important. U.S.–China relations have never been made only by presidents, diplomats and official communiques. They have also been shaped by athletes, musicians, restaurant owners, students and tourists.

The basketball reference was especially resonant. Sports have long offered a softer language for U.S.–China relations. In April 2026, just weeks before Trump’s visit, China and the U.S. marked the 55th anniversary of ping-pong diplomacy — the famous 1971 exchange in which a “little ball” helped move the “big ball” of world politics.

Basketball now plays a similar role. For many Chinese fans, the NBA is a deeply familiar world of players, teams and memories that represents the spirit of America: Michael Jordan, Kobe Bryant, LeBron James and Yao Ming. That reservoir of affection has survived even periods of political tension. Trump, in invoking it, was drawing on something real.

A second act in the US?

The main lesson of all this symbolism is that, in U.S.–China relations, atmosphere has never been secondary.

Diplomatic theater cannot settle disputes over technology or Taiwan, or determine the future of the global order. But it can shape the mood in which rivalries are managed, and the stories that leaders tell their public about what the relationship means.

And on that front, the summit worked on several levels. To the Chinese audience, it presented their leaders as confident and capable of managing a tense relationship with the U.S. on China’s own cultural terms.

For Trump and the American delegation, it offered a lesson in Chinese traditions and culture that promotes deeper understanding across political divides. And for both societies, the references for food, sports and popular culture created a more neutral ground on which connection could still be imagined.

From the 1970s opening to Trump’s 2017 visit to the Forbidden City, and from the Temple of Heaven photo-op to the walk among old trees at Zhongnanhai in 2026, cultural staging remains central to how China presents itself to America — and how America is invited to imagine China. It was announced on May 15 that Xi will pay a state visit to the U.S. in September at the invitation of Trump. If that happens, the theater of diplomacy will move to American soil, and the question will be how Washington chooses to stage China in return.

Xianda Huang does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

Diplomacy often masquerades as theater. And nearly nine years after his first state visit to China, Donald Trump returned to Beijing with an extended cast of characters.

Alongside the U.S. president on his May 2026 visit was a senior delegation of politicians including his secretary of defense, and a phalanx of business leaders and technology executives. It was a traveling display of American political and corporate power.

Not that the hosting Chinese were short of symbolic gestures themselves. Trump’s first China visit in 2017 had already shown how far Beijing was willing to go to turn diplomacy into theater. On that occasion, Chinese President Xi Jinping and his wife Peng Liyuan personally accompanied Donald and Melania Trump through the Forbidden City, Beijing’s former imperial palace, drinking tea inside the palace walls and taking in a Peking opera at the Belvedere of Pleasant Sounds, a Qing imperial theater built for court entertainment.

So what was being conveyed this time around? As a cultural historian of modern China, I took a peek beyond the official statements and trade headlines of the Xi-Trump summit and into the images, gestures and cultural symbolism on display.

The weight of heaven

The formal choreography began at Beijing’s Great Hall of the People, where the two leaders exchanged views on the Iran conflict, the war in Ukraine and the Korean Peninsula, among other items.

But the more interesting story of the visit, to me, was told outside the meeting room.

After their two-hour bilateral meeting, Trump and Xi paid a cultural visit to the Temple of Heaven in Southern Beijing. Built in the early 15th century, the temple is China’s most complete surviving imperial religious complex. For nearly five centuries, emperors of the Ming and Qing dynasties came here to worship Heaven and pray for good harvests.

Its most recognizable structure, the Hall of Prayer for Good Harvests, rises in three tiers of blue-glazed tiles above a marble platform, its circular form and crimson columns translating cosmology into architecture. UNESCO inscribed the site as a World Heritage Site in 1998, recognizing it as “a masterpiece of architecture and landscape design.”

When Trump and Xi posed for photographs, they were standing in a place long associated with cosmic order and the welfare of the people. To bring a foreign leader there is to invite a particular reading of the relationship: not simply as a bargain between states, but as a relationship that Beijing hopes to associate with order, abundance and peace.

There was also a more practical layer to this symbolism. The Temple of Heaven links political authority to agricultural abundance. Emperors came here to pray not for abstract harmony but for grain. That made it a pointed setting for a visit in which American agricultural exports — soybeans, grains and beef among them — were expected to matter.

For Trump, any Chinese commitment to buy more U.S. farm goods would have clear domestic political value. For Xi, the setting allowed a hard bargaining issue — farm purchases — to be translated into an older symbolic language of harvest that spoke to both domestic and international audiences.

Before Trump, Kissinger

Trump was not the first American statesman to be brought to the Temple of Heaven.

In July 1971, Henry Kissinger, then national security adviser to President Richard Nixon, arrived in Beijing on his famous secret mission — the back-channel visit that helped re-open the door between two countries that had little direct contact for more than two decades. Between tense negotiations with Chinese premier Zhou Enlai, Kissinger made time to visit the temple.

There, standing amid the old cypress groves, he was said to have been deeply moved by the timeless atmosphere of the hall and its surroundings.

The motif of old trees and deep time returned on May 15, when Xi gave Trump a rare walk through Zhongnanhai, the walled compound that now houses the core of China’s party-state leadership. Reuters reported that a hot mic captured Xi drawing Trump’s attention to the age of the trees around them — some centuries old, some said to be more than a thousand years old. When Trump asked whether Xi had taken other presidents on similar walks, Xi replied that he had only rarely.

Together, the Kissinger anecdote and the Zhongnanhai walk reveal a recurring logic in Chinese-American diplomacy: America’s fast-moving economy is invited to look at China’s sense of tradition. Xi has used this tactic with other leaders, too. When French President Emmanuel Macron visited China in 2023, he attended a guqin performance invoking the classical idea of the zhiyin — the rare listener who truly understands one’s music.

Basketball and roast duck

Trump’s visit was not staged only through imperial grandeur, however. It also moved into a more familiar register: food, sports and popular culture.

The state dinner on May 14 was another study in careful hospitality. Chefs designed the menu to honor both Chinese culinary prestige and Americans’ — and Trump’s — known preferences: Peking roast duck, crispy beef ribs, pan-fried pork bun, tiramisu and fruit and ice cream.

Trump thanked Xi for a “magnificent welcome like none other,” then replied in a language more recognizably his own. He spoke not only of power politics but of people-to-people ties: Chinese workers who helped build America’s railroads, Chinese enthusiasm for basketball and blue jeans and the sheer presence of Chinese restaurants across the U.S.

The examples were characteristically Trumpian — simple, vivid and easy to grasp. But they pointed to something important. U.S.–China relations have never been made only by presidents, diplomats and official communiques. They have also been shaped by athletes, musicians, restaurant owners, students and tourists.

The basketball reference was especially resonant. Sports have long offered a softer language for U.S.–China relations. In April 2026, just weeks before Trump’s visit, China and the U.S. marked the 55th anniversary of ping-pong diplomacy — the famous 1971 exchange in which a “little ball” helped move the “big ball” of world politics.

Basketball now plays a similar role. For many Chinese fans, the NBA is a deeply familiar world of players, teams and memories that represents the spirit of America: Michael Jordan, Kobe Bryant, LeBron James and Yao Ming. That reservoir of affection has survived even periods of political tension. Trump, in invoking it, was drawing on something real.

A second act in the US?

The main lesson of all this symbolism is that, in U.S.–China relations, atmosphere has never been secondary.

Diplomatic theater cannot settle disputes over technology or Taiwan, or determine the future of the global order. But it can shape the mood in which rivalries are managed, and the stories that leaders tell their public about what the relationship means.

And on that front, the summit worked on several levels. To the Chinese audience, it presented their leaders as confident and capable of managing a tense relationship with the U.S. on China’s own cultural terms.

For Trump and the American delegation, it offered a lesson in Chinese traditions and culture that promotes deeper understanding across political divides. And for both societies, the references for food, sports and popular culture created a more neutral ground on which connection could still be imagined.

From the 1970s opening to Trump’s 2017 visit to the Forbidden City, and from the Temple of Heaven photo-op to the walk among old trees at Zhongnanhai in 2026, cultural staging remains central to how China presents itself to America — and how America is invited to imagine China. It was announced on May 15 that Xi will pay a state visit to the U.S. in September at the invitation of Trump. If that happens, the theater of diplomacy will move to American soil, and the question will be how Washington chooses to stage China in return.

Xianda Huang does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

The nonprofit ARC Prize Foundation on May 1, 2026, released the results of a new benchmark: a test of an AI system’s ability to solve a game. The results were striking – humans scored 100%, while the most advanced AI systems scored under 1%.

At first glance, this may be surprising to users of AI who are impressed by its polished essays, codebases and multistep projects generated in seconds. How can these brilliant AI systems struggle with these simple Tetris-shape puzzles?

That confusion points to a risk: AI is becoming integrated into everyday life faster than people can make sense of it.

We are cognitive psychologists who study how to teach difficult concepts. To recognize the limits and risks of today’s AI agent systems, it’s important for people to grasp that the systems can both accomplish superhuman feats and make mistakes few humans would. To that end, we propose a new way to think about AIs: as button-pushing explorers.

Mental models for AI

We teach college students, a group rapidly incorporating AI tools into their daily routines. That gives us regular opportunities to ask what they think is going on with AI. The answers vary widely. One student said that someone at OpenAI or Anthropic is reading and approving every response the system generates. Another, more succinctly, said, “It’s magic.”

These responses illustrate two tempting ways of making sense of AI. At one extreme, AI is treated as an inscrutable black box – a powerful but ultimately mysterious force. At another, people explain it using the same assumptions they use to understand other humans: that its outputs reflect reasoning or judgment.

The worry is that these misinterpretations don’t go away as users gain more experience interacting with AI, and they might get reinforced. When AI performs well, its output can feel like evidence of understanding or confirmation that it really is something like magic. That apparent success makes it harder to question what the system is actually doing. Biases can seem logical or inevitable; harmful behavior can look like a deliberate choice or even fate, as if it could not have gone any other way.

Saying that AI models are shaped by patterns in data, training processes and system design is true, but that’s too abstract to tell people when to trust the systems’ outputs or when they might fail. To help people avoid misplaced trust in AI, AI literacy efforts will need to include some mechanistic understanding of what produces their behavior – explanations that are perhaps not perfectly accurate but useful. Statistician George Box once wrote, “All models are wrong, but some are useful.”

Researchers have come up with several mental models for large language models. One is “stochastic parrot,” which shows that the models use statistical methods – stochastic refers to probabilities – to mimic responses with no understanding of meaning. Another is “bag of words,” which emphasizes that the models are collections of words – for example, all English words found on the internet – with a mechanism for giving you the best set of words based on your prompt.

These ways of thinking about large language models were never meant to be complete accounts of the systems. But the metaphors serve an important cognitive purpose: They push back against the idea that fluent language is necessarily caused by humanlike understanding.

But as the AI systems people use are increasingly powerful agents capable of stringing together actions on their own, it’s important for people to have a different kind of mental model: one that explains how they act. One place to find such a model is in earlier research on AI systems that learned to play Atari 2600 games. These systems didn’t understand the games the way humans do, but they still managed to rack up a lot of points.

The simple loop: Act, observe, adjust

Imagine a neural network, a relatively simple kind of AI model, placed into a video game it has never seen before. It does not “understand” the game like a human would. It has no idea whether it’s shooting space invaders or navigating an ancient pyramid. It doesn’t know the goals or rules.

Instead, it learns to play through a simple loop: Take an action – move left, jump, shoot – observe what changes, and then adjust. If an action leads to a good outcome, such as gaining points, it adjusts to become more likely to take similar actions in similar situations. If it leads to a bad outcome, such as losing a life, it adjusts in the opposite direction.

Even this simple mechanism can produce surprisingly capable behavior. Over time, by repeating this loop, the neural networks learned to play a wide range of Atari games – but not all games.

There is one game that famously stumped these early neural networks: Montezuma’s Revenge. To make progress, a player must carry out a long sequence of actions – climbing ladders, avoiding obstacles, retrieving keys – before receiving any reward at all. Unlike simpler games, most actions offer very little immediate feedback. The game required something like goal-directed, long-term planning.

Early neural networks would try a few actions, receive no reward and fail to make further progress through Montezuma’s underground pyramid. From the system’s perspective, all actions looked equally useless. But researchers made a breakthrough by changing the feedback signal. Instead of rewarding only success, they also rewarded the system for doing something new. The rewards were for visiting parts of the game it had not seen before or trying actions it had not previously taken. This tweak encouraged exploration.

With that change, performance improved dramatically. The neural network began navigating obstacles, taking multiple steps toward goals and adapting when things went wrong. From the outside, this kind of behavior can look like planning or problem-solving. But what looks like planning was not caused by sophisticated planning abilities. The underlying mechanism is still the same simple loop: act, observe, adjust.

This kind of system isn’t a stochastic parrot or a bag of words. It’s closer to a button-pushing explorer: something that doesn’t understand the world in a human sense but moves forward by pushing buttons, seeing what happens and adjusting what it does next.

From video games to modern AI agents

Today’s AI systems can do far more than play games like Montezuma’s Revenge. They can coordinate tools, write and run code, and carry out multistep projects. The range of possible actions is much larger, and the environments in which they operate are increasingly complex.

But these agents are still fundamentally button-pushing explorers. The behavior can be sophisticated, but the process that produces it is not. Humans can often infer how a new environment works after just a few observations. Systems that rely on these feedback loops cannot. They need to try many actions and see what happens before they can make progress.

This helps explain both the strengths of these AI systems and some of their most concerning failures. What these agents learn depends on what is being rewarded. And in real-world systems, those reward signals are often imperfect.

AI systems that conduct negotiations aim to maximize their client’s interests, sometimes with deceptive tactics. Rental pricing software used by landlords ends up price fixing. Marketing tools generate persuasive but misleading reviews.

These systems aren’t trying to be evil or greedy. They are adjusting to the signals they are given. From the button-pushing explorer perspective, these failures are downright predictable.

Effective AI literacy means holding two ideas at once: These systems can do surprisingly complex things, and they are not doing them the way humans do. If AI is seen as humanlike or magical, its outputs feel authoritative. But if it is understood, even imperfectly, as a button-pushing explorer shaped by feedback, people are likely to ask better questions: Why is it doing this? What shaped this behavior? What might it be missing?

That’s the difference between being impressed by AI and being able to reason about it.

Ji Y. Son receives research funding from the Gates Foundation and Valhalla Foundation.

Alice Xu does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

If you follow the headlines, it can seem like K-12 schools in the United States are a political battlefield.

Some conservative parents and advocacy groups are lobbying to remove certain books from classrooms and libraries, most often those that highlight LGBTQ+ issues or race and racism.

Some civil liberties groups, librarians and progressive parents, meanwhile, are pushing back against book bans, saying they are a form of unnecessary censorship.

Parents and school boards also are clashing over a range of other issues, ranging from how transgender and nonbinary students are treated and which bathrooms they can use, to whether teachers should use artificial intelligence in the classroom.

Beyond this evidence of political polarization, though, there’s another, less divisive reality. Ask people to name their best teacher, and regardless of their political affiliation, they will likely offer a similar answer. Most people will say that they learned a lot from a teacher who knew them, cared about them and made learning relevant to their lives.

Over five years, from 2020 through 2025, we asked more than 2,000 Americans, including Democrats, Republicans and independents, what makes a very good teacher. We expected deep partisan divides. Instead, we found something rare: genuine, cross-partisan agreement.

How we ran the study

We began in 2020 with a nationally representative survey of 334 adults, asking them to recall a teacher they learned a lot from. We then asked the survey participants to look at 10 statements that might describe a good teacher and rank them from most to least important.

Five of the statements we offered focused on relationships – like caring about students, making educational lessons relevant and giving students individualized support. The other five focused on whether teachers covered a lot of material, rewarded top performers with grades or prizes, and whether they applied rules consistently to all students.

Respondents generally focused on highlighting the same seven out of 10 statements, giving us a vision of how they perceived a very good teacher. People prioritized the same factors – how much the teachers cared about their students and whether they supported them – regardless of their age, race, gender or political affiliation. Republicans and Democrats were indistinguishable in their descriptions of effective teaching.

People did not prioritize whether teachers covered a lot of material, made students compete or ran a strict and disciplined classroom.

In 2022, we conducted a similar survey of 179 teachers in Arizona and California. The results echoed our 2020 survey participants’ view: Teachers also defined very good teachers as ones who emphasized relationships, made lessons relevant and knew the subject matter.

Given the prominence of politically charged education debates, we were a bit surprised by our results. We began to wonder: Do people privately agree on what it means to be a good teacher, but change their opinion if their image of good teaching is associated with an ideological orientation they disagree with?

Adding a partisan label

To explore this question in late 2024 and early 2025, we ran a third experiment with a nationally representative sample of 1,562 adults from a range of political backgrounds.

We gave all participants the same description of a very good teacher, identified in our previous experiments. We then randomly noted if these descriptions of a good teacher were endorsed by Democrats, Republicans or people with no political affiliation.

When the participants read the teacher descriptions without any political labels attached, about 85% of Democrats, Republicans and independents agreed with the description of a very good teacher.

When we added a note saying that a political party the survey participant did not identify endorsed a particular description of a good teacher, they became less likely to support the statement.

The effect was sharpest among Republicans: Support fell from 85% to 64% when the description was tied to Democrats. Democrats’ agreement slipped less, from 86% to 76%, when the description was tied to Republicans.

Even with these caveats, nearly two-thirds of Republicans and Democrats still agreed on what it means to be a good teacher.

Political scientists call this affective polarization: How we react to an idea depends not just on the idea, but on who we think supports it.

At the national level, education is often framed as an intractable partisan conflict.

Yet at the individual level, many Americans continue to express confidence in their own local schools. Our findings suggest that part of this gap may be driven by how issues are framed rather than by fundamentally incompatible beliefs.

This matters more than you might think

Federal and state education policy over the past four decades, including laws like No Child Left Behind, which mandated routine federal testing in reading and math, emphasize testing and competition. These priorities don’t always match what Americans across the political spectrum say they value most.

Americans continue to differ on many important education questions, including what children should learn in school, the role of school boards and other issues.

But these disagreements coexist with a shared beliefs about what good teaching looks like in practice.

Recognizing this gap could open new possibilities for education reform. When debates focus exclusively on disagreements, they can obscure areas of agreement that might otherwise serve as starting points for collaboration.

We encourage readers to go ahead and run a similar, small experiment: Ask people about their best teacher, then listen to what they say. The answer, it turns out, is likely more unifying than you expect.

For this specific project, Gustavo E. Fischman received funding from the Institute of Social Science Research at ASU. He also received funds for other projects from the National Science Foundation, the Spencer Foundation, the Open Society Foundation, the IDRC, and the Fulbright Commission.

Eric Haas and Margarita Pivovarova do not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.

The nonprofit ARC Prize Foundation on May 1, 2026, released the results of a new benchmark: a test of an AI system’s ability to solve a game. The results were striking – humans scored 100%, while the most advanced AI systems scored under 1%.

At first glance, this may be surprising to users of AI who are impressed by its polished essays, codebases and multistep projects generated in seconds. How can these brilliant AI systems struggle with these simple Tetris-shape puzzles?

That confusion points to a risk: AI is becoming integrated into everyday life faster than people can make sense of it.

We are cognitive psychologists who study how to teach difficult concepts. To recognize the limits and risks of today’s AI agent systems, it’s important for people to grasp that the systems can both accomplish superhuman feats and make mistakes few humans would. To that end, we propose a new way to think about AIs: as button-pushing explorers.

Mental models for AI

We teach college students, a group rapidly incorporating AI tools into their daily routines. That gives us regular opportunities to ask what they think is going on with AI. The answers vary widely. One student said that someone at OpenAI or Anthropic is reading and approving every response the system generates. Another, more succinctly, said, “It’s magic.”

These responses illustrate two tempting ways of making sense of AI. At one extreme, AI is treated as an inscrutable black box – a powerful but ultimately mysterious force. At another, people explain it using the same assumptions they use to understand other humans: that its outputs reflect reasoning or judgment.

The worry is that these misinterpretations don’t go away as users gain more experience interacting with AI, and they might get reinforced. When AI performs well, its output can feel like evidence of understanding or confirmation that it really is something like magic. That apparent success makes it harder to question what the system is actually doing. Biases can seem logical or inevitable; harmful behavior can look like a deliberate choice or even fate, as if it could not have gone any other way.

Saying that AI models are shaped by patterns in data, training processes and system design is true, but that’s too abstract to tell people when to trust the systems’ outputs or when they might fail. To help people avoid misplaced trust in AI, AI literacy efforts will need to include some mechanistic understanding of what produces their behavior – explanations that are perhaps not perfectly accurate but useful. Statistician George Box once wrote, “All models are wrong, but some are useful.”

Researchers have come up with several mental models for large language models. One is “stochastic parrot,” which shows that the models use statistical methods – stochastic refers to probabilities – to mimic responses with no understanding of meaning. Another is “bag of words,” which emphasizes that the models are collections of words – for example, all English words found on the internet – with a mechanism for giving you the best set of words based on your prompt.

These ways of thinking about large language models were never meant to be complete accounts of the systems. But the metaphors serve an important cognitive purpose: They push back against the idea that fluent language is necessarily caused by humanlike understanding.

But as the AI systems people use are increasingly powerful agents capable of stringing together actions on their own, it’s important for people to have a different kind of mental model: one that explains how they act. One place to find such a model is in earlier research on AI systems that learned to play Atari 2600 games. These systems didn’t understand the games the way humans do, but they still managed to rack up a lot of points.

The simple loop: Act, observe, adjust

Imagine a neural network, a relatively simple kind of AI model, placed into a video game it has never seen before. It does not “understand” the game like a human would. It has no idea whether it’s shooting space invaders or navigating an ancient pyramid. It doesn’t know the goals or rules.

Instead, it learns to play through a simple loop: Take an action – move left, jump, shoot – observe what changes, and then adjust. If an action leads to a good outcome, such as gaining points, it adjusts to become more likely to take similar actions in similar situations. If it leads to a bad outcome, such as losing a life, it adjusts in the opposite direction.

Even this simple mechanism can produce surprisingly capable behavior. Over time, by repeating this loop, the neural networks learned to play a wide range of Atari games – but not all games.

There is one game that famously stumped these early neural networks: Montezuma’s Revenge. To make progress, a player must carry out a long sequence of actions – climbing ladders, avoiding obstacles, retrieving keys – before receiving any reward at all. Unlike simpler games, most actions offer very little immediate feedback. The game required something like goal-directed, long-term planning.

Early neural networks would try a few actions, receive no reward and fail to make further progress through Montezuma’s underground pyramid. From the system’s perspective, all actions looked equally useless. But researchers made a breakthrough by changing the feedback signal. Instead of rewarding only success, they also rewarded the system for doing something new. The rewards were for visiting parts of the game it had not seen before or trying actions it had not previously taken. This tweak encouraged exploration.

With that change, performance improved dramatically. The neural network began navigating obstacles, taking multiple steps toward goals and adapting when things went wrong. From the outside, this kind of behavior can look like planning or problem-solving. But what looks like planning was not caused by sophisticated planning abilities. The underlying mechanism is still the same simple loop: act, observe, adjust.

This kind of system isn’t a stochastic parrot or a bag of words. It’s closer to a button-pushing explorer: something that doesn’t understand the world in a human sense but moves forward by pushing buttons, seeing what happens and adjusting what it does next.

From video games to modern AI agents

Today’s AI systems can do far more than play games like Montezuma’s Revenge. They can coordinate tools, write and run code, and carry out multistep projects. The range of possible actions is much larger, and the environments in which they operate are increasingly complex.

But these agents are still fundamentally button-pushing explorers. The behavior can be sophisticated, but the process that produces it is not. Humans can often infer how a new environment works after just a few observations. Systems that rely on these feedback loops cannot. They need to try many actions and see what happens before they can make progress.

This helps explain both the strengths of these AI systems and some of their most concerning failures. What these agents learn depends on what is being rewarded. And in real-world systems, those reward signals are often imperfect.

AI systems that conduct negotiations aim to maximize their client’s interests, sometimes with deceptive tactics. Rental pricing software used by landlords ends up price fixing. Marketing tools generate persuasive but misleading reviews.

These systems aren’t trying to be evil or greedy. They are adjusting to the signals they are given. From the button-pushing explorer perspective, these failures are downright predictable.

Effective AI literacy means holding two ideas at once: These systems can do surprisingly complex things, and they are not doing them the way humans do. If AI is seen as humanlike or magical, its outputs feel authoritative. But if it is understood, even imperfectly, as a button-pushing explorer shaped by feedback, people are likely to ask better questions: Why is it doing this? What shaped this behavior? What might it be missing?

That’s the difference between being impressed by AI and being able to reason about it.

Ji Y. Son receives research funding from the Gates Foundation and Valhalla Foundation.

Alice Xu does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

If you follow the headlines, it can seem like K-12 schools in the United States are a political battlefield.

Some conservative parents and advocacy groups are lobbying to remove certain books from classrooms and libraries, most often those that highlight LGBTQ+ issues or race and racism.

Some civil liberties groups, librarians and progressive parents, meanwhile, are pushing back against book bans, saying they are a form of unnecessary censorship.

Parents and school boards also are clashing over a range of other issues, ranging from how transgender and nonbinary students are treated and which bathrooms they can use, to whether teachers should use artificial intelligence in the classroom.

Beyond this evidence of political polarization, though, there’s another, less divisive reality. Ask people to name their best teacher, and regardless of their political affiliation, they will likely offer a similar answer. Most people will say that they learned a lot from a teacher who knew them, cared about them and made learning relevant to their lives.

Over five years, from 2020 through 2025, we asked more than 2,000 Americans, including Democrats, Republicans and independents, what makes a very good teacher. We expected deep partisan divides. Instead, we found something rare: genuine, cross-partisan agreement.

How we ran the study

We began in 2020 with a nationally representative survey of 334 adults, asking them to recall a teacher they learned a lot from. We then asked the survey participants to look at 10 statements that might describe a good teacher and rank them from most to least important.

Five of the statements we offered focused on relationships – like caring about students, making educational lessons relevant and giving students individualized support. The other five focused on whether teachers covered a lot of material, rewarded top performers with grades or prizes, and whether they applied rules consistently to all students.

Respondents generally focused on highlighting the same seven out of 10 statements, giving us a vision of how they perceived a very good teacher. People prioritized the same factors – how much the teachers cared about their students and whether they supported them – regardless of their age, race, gender or political affiliation. Republicans and Democrats were indistinguishable in their descriptions of effective teaching.

People did not prioritize whether teachers covered a lot of material, made students compete or ran a strict and disciplined classroom.

In 2022, we conducted a similar survey of 179 teachers in Arizona and California. The results echoed our 2020 survey participants’ view: Teachers also defined very good teachers as ones who emphasized relationships, made lessons relevant and knew the subject matter.

Given the prominence of politically charged education debates, we were a bit surprised by our results. We began to wonder: Do people privately agree on what it means to be a good teacher, but change their opinion if their image of good teaching is associated with an ideological orientation they disagree with?

Adding a partisan label

To explore this question in late 2024 and early 2025, we ran a third experiment with a nationally representative sample of 1,562 adults from a range of political backgrounds.

We gave all participants the same description of a very good teacher, identified in our previous experiments. We then randomly noted if these descriptions of a good teacher were endorsed by Democrats, Republicans or people with no political affiliation.

When the participants read the teacher descriptions without any political labels attached, about 85% of Democrats, Republicans and independents agreed with the description of a very good teacher.

When we added a note saying that a political party the survey participant did not identify endorsed a particular description of a good teacher, they became less likely to support the statement.

The effect was sharpest among Republicans: Support fell from 85% to 64% when the description was tied to Democrats. Democrats’ agreement slipped less, from 86% to 76%, when the description was tied to Republicans.

Even with these caveats, nearly two-thirds of Republicans and Democrats still agreed on what it means to be a good teacher.

Political scientists call this affective polarization: How we react to an idea depends not just on the idea, but on who we think supports it.

At the national level, education is often framed as an intractable partisan conflict.

Yet at the individual level, many Americans continue to express confidence in their own local schools. Our findings suggest that part of this gap may be driven by how issues are framed rather than by fundamentally incompatible beliefs.

This matters more than you might think

Federal and state education policy over the past four decades, including laws like No Child Left Behind, which mandated routine federal testing in reading and math, emphasize testing and competition. These priorities don’t always match what Americans across the political spectrum say they value most.

Americans continue to differ on many important education questions, including what children should learn in school, the role of school boards and other issues.

But these disagreements coexist with a shared beliefs about what good teaching looks like in practice.

Recognizing this gap could open new possibilities for education reform. When debates focus exclusively on disagreements, they can obscure areas of agreement that might otherwise serve as starting points for collaboration.

We encourage readers to go ahead and run a similar, small experiment: Ask people about their best teacher, then listen to what they say. The answer, it turns out, is likely more unifying than you expect.

For this specific project, Gustavo E. Fischman received funding from the Institute of Social Science Research at ASU. He also received funds for other projects from the National Science Foundation, the Spencer Foundation, the Open Society Foundation, the IDRC, and the Fulbright Commission.

Eric Haas and Margarita Pivovarova do not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.