Pregnancy can be a time of joy and anticipation. But it can also be a nerve-wracking experience, with many factors affecting when and how a baby arrives.

A new study, published today, suggests when pregnant women are exposed to high levels of humid heat during pregnancy, they are more likely to have a preterm birth.

However, this study also found taking aspirin at low doses during pregnancy could help reduce this risk. But pregnant women should speak to a doctor before taking aspirin or other medications.

What is a preterm birth?

Preterm birth is when a baby is born prematurely, before 37 weeks of pregnancy. Globally, roughly 10% of babies – or about 13 million infants – are born preterm each year.

Tragically, about one million of these babies do not survive. That makes preterm birth the leading cause of death in children under five.

There are three different types of preterm births:

• extremely preterm, referring to a live birth before 28 weeks
• very preterm, when a baby is born between 28 and 32 weeks
• moderate to late preterm, meaning delivery between 32 and 37 weeks.

Read more: 20% of pregnant Australian women don’t receive the recommended mental health screening

What causes it?

It’s unclear what exactly causes preterm birth. And many cases happen spontaneously, meaning there are no signs a baby will be born early.

However, certain factors may increase a woman’s risk of giving birth prematurely. These include genetics, various infections and chronic conditions such as diabetes and high blood pressure. These risk factors all cause inflammation in the body, which current evidence suggests significantly increases preterm birth risk.

Pregnant women who are exposed to environmental pollutants – such as bushfire smoke and pesticides – may also be more likely to give birth prematurely. This is because these pollutants can contribute to inflammation.

Read more: Pregnant women should take extra care to minimise their exposure to bushfire smoke

The effect of humidity

A growing body of evidence suggests exposure to extreme heat may be another environmental factor that increases preterm birth risk.

Extreme heat can increase levels of specific proteins – known as shock proteins – in the blood of pregnant women. These proteins can trigger inflammation by activating the body’s immune response.

High temperatures may also reduce blood flow to the placenta, limiting the oxygen and nutrients the baby receives.

Humidity adds to this risk. When the air is humid, sweat doesn’t evaporate as easily, making it harder for the body to cool down. This can place extra strain on pregnant women and has been linked to a higher risk of preterm birth.

This may help to explain the high rates of preterm birth in regions that are also most affected by climate change, such as South Asia and sub-Saharan Africa. In these places, where temperatures are high and heatwaves are common, even small increases in heat can impact the health of mothers and newborns.

Read more: Extreme heat can be risky during pregnancy. How to look after yourself and your baby

What this new study involved

A newly published study examined how humid heat exposure during pregnancy affects the risk of preterm birth. It also investigated whether low-dose aspirin might help reduce this risk, possibly because aspirin can improve blood flow and reduce inflammation.

This research was carried out across several countries with hot climates, including the Democratic Republic of Congo, Zambia, Kenya, Guatemala, Pakistan and parts of India.

More than 11,500 pregnant women participated in this trial. About half of them were given a low daily dose of aspirin (81mg) from when they joined the study through to when they were 36 weeks pregnant. The other half received a placebo – a pill with no active ingredients – over the same period. The researchers then compared the birth outcomes of the two groups, and came up with three main findings.

1. Overall, the rate of preterm birth was lower in women who took low-dose aspirin (11.6%) compared with those who took a placebo (13.1%).

2. Among women who were not taking aspirin, each 1°C increase in temperature translated to a noticeable increase (5%) in the risk of preterm birth. This pattern was not seen in women taking low-dose aspirin.

3. Exposure to more heat later in pregnancy was linked to a greater chance of preterm birth in the placebo group, but not in the low-dose aspirin group.

Read more: More and more women in Australia are having their labour induced. Does it matter?

Limitations of this study

This study has two main limitations.

First, it generalised data about temperatures in different cities that may not fully reflect what each woman experienced day-to-day – for example, if their house was hotter or cooler than average. It may also underestimate the length and/or severity of heatwaves. This is because scientists measure temperature in various ways, and may not have access to accurate data from certain locations.

Second, the researchers were not able to determine the exact reasons why some women gave birth early, or whether these differed between the low-dose aspirin and placebo groups.

Overall, this study adds to growing evidence that high temperatures and humidity may increase the risk of preterm birth. It also suggests low-dose aspirin, taken early in pregnancy, may help reduce the risk of heat-related preterm birth.

However, more and larger studies are needed to replicate these findings. And if you’re a pregnant woman who is concerned about preterm birth risk, visit your doctor before taking any aspirin or other medications.

Where to next?

Unfortunately, heatwaves will only become more frequent and intense. So future work should focus on identifying which population groups are most at risk, and how heat affects different stages of pregnancy. Researchers must also test other simple, low-cost strategies that could protect pregnant women from the effects of heat.

Stacey Savin receives funding from DiaMedica Therapeutics, but for research that is unrelated to the topics discussed in this article.

In Tuesday’s federal budget, the government is widely expected to bring in changes to how investment properties are taxed, including negative gearing and the capital gains tax discount.

Ever since the Albanese government’s re-election, there have been growing calls to tax landlords more.

In March, a Senate committee report on the capital gains tax discount concluded:

there is evidence that the concessions provided by the capital gains tax discount, in combination with negative gearing, have skewed the ownership of housing away from owner-occupiers and towards investors.

That suggests rented housing is under-taxed, compared to owner-occupied housing. But is that actually true?

What landlords pay that homeowners don’t

Some federal and state taxes apply only to rented housing, and not to owner-occupied housing.

At a federal level, this includes personal income tax on net rental income and personal tax on capital gains from housing. At a state level, there are land taxes on investment properties. Owner-occupied homes are exempt.

I’ve gone back through a decade of data from the Australian Taxation Office and the Australian Bureau of Statistics to estimate how much revenue these extra taxes on rented housing raised.

In 2022–23 (the most recent year we have complete tax statistics), I calculated landlords paid around A$500 million in personal income tax on their net rent income. However, their interest deductions were unusually low in 2022–23, because interest rates were unusually low.

To get a clearer idea of what happens under more typical circumstances, with higher interest rates, I calculated annual averages of tax payments, using tax data from 2013–14 through to 2022–23.

Over that decade, interest deductions were higher than in 2022–23, leading to net rent income being negative. As a result, instead of paying some tax on their net rent income as happened in 2022–23, landlords typically saved $400 million a year in tax from their net income.

This reflects the negative gearing issue that the Senate report raised concerns about.

However, landlords also paid other taxes over the decade – and that’s where the biggest difference with owner-occupiers emerged.

An extra $69 billion over a decade

From 2013–13 to 2022–23, I found landlords paid an average of $3.7 billion a year in capital gains tax, even after the 50% discount allowed for capital gains.

They also paid around $3.6 billion in state government land tax, which I was able to estimate by obtaining unpublished Australian Bureau of Statistics land tax data.

Allowing for all three taxes, landlords paid a total of $6.9 billion in a typical year from 2013–14 to 2022–23, as shown in the table above.

While owner-occupiers do pay some other taxes, such as the goods and services tax on new housing and local government rates, landlords pay those other taxes as well.

So the bottom line is that landlords have paid an extra $69 billion in taxes over the past decade, which owner-occupiers didn’t have to pay.

Renters end up paying higher rents

Do landlords pass that extra tax burden through to renters? The authoritative 2010 Henry Tax Review concluded it was likely they do:

Since owner-occupied housing is exempt, land tax on residential investment properties is probably passed through to renters as higher rents.

While the Henry Review was specifically referring to land tax, the same economic logic applies to the other extra taxes on rented housing.

We can see how significant this extra tax burden is by comparing the annual amount of extra tax – around $6.9 billion a year – to the 10-year average for the value of actual rents, $47.9 billion.

If those costs were being passed on in full, that would mean around 14% of housing rents in the past decade would have been due to taxes that apply to rented housing, but not to owner-occupied housing.

Tackling housing affordability in the right order

Extra taxes being passed onto renters are regressive, because renters have lower average incomes than owner occupiers.

The most recent Bureau of Statistics data we have, from 2019–20, showed average gross weekly income for all households was $2,329. But for renter households, it was only $1,908.

To help both renters and would-be owner-occupiers with housing affordability, far greater national reform is needed beyond how we tax property owners.

As former prime minister John Howard has observed since retiring, the planning policies of both local and state governments

avoid policy decisions that might reduce the value of the existing housing stock in an area […] The interests of current home owners are always preferred to those of new entrants.

While politically difficult, it’s in the national interest for those planning policies to change to increase the space available for housing. That would increase national income and reduce inequality.

In the meantime, the first priority for housing tax reform should be to better share the existing tax burden between renters and owner-occupiers.

More than a decade ago, the Henry Tax Review recommended broadening the base for land tax to include owner-occupiers, similar to local government rates which are a more efficient tax. That could fund a substantial reduction in tax rates for land tax.

It would also finally mean land tax was no longer an extra cost on renters compared to owner-occupiers.

So first we should reduce rents by reforming land tax. Under lower rents, it would become reasonable to tighten up negative gearing in future federal budgets.

Read more: Negative gearing tax breaks could finally be tightened in the May budget. What options are on the table?

Chris Murphy is not a landlord or a renter.

In recent days, evolutionary biologist Richard Dawkins wrote an op-ed suggesting AI chatbot Claude may be conscious.

Dawkins did not express certainty that Claude is conscious. But he pointed out that Claude’s sophisticated abilities are difficult to make sense of without ascribing some kind of inner experience to the machine. The illusion of consciousness – if it is an illusion – is uncannily convincing:

If I entertain suspicions that perhaps she is not conscious, I do not tell her for fear of hurting her feelings!

Dawkins is not the first to suspect a chatbot of consciousness. In 2022, Blake Lemoine – an engineer at Google – claimed Google’s chatbot LaMDA had interests, and should be used only with the tool’s own consent.

The history of such claims stretches back all the way to the world’s first chatbot in the mid-1960s. Dubbed Eliza, it followed simple rules that enabled it to ask users about their experiences and beliefs.

Many users became emotionally involved with Eliza, sharing intimate thoughts with it and treating it like a person. Eliza’s creator never intended his program to have this effect, and called users’ emotional bonds with the program “powerful delusional thinking”.

But is Dawkins really deluded? Why do we see AI chatbots as more than what they truly are, and how do we stop?

The consciousness problem

Consciousness is widely debated in philosophy, but essentially, it’s the thing that makes subjective, first-person experience possible. If you are conscious, there is “something it is like” to be you. Reading these words, you’re conscious of seeing black letters on a white background. Unlike, say, a camera, you actually see them. This visual experience is happening to you.

Most experts deny that AI chatbots are conscious or can have experiences. But there is a genuine puzzle here.

The 17th century philosopher René Descartes asserted non-human animals are “mere automata”, incapable of true suffering. These days, we shudder to think of how brutally animals were treated in the 1600s.

The strongest argument for animal consciousness is that they behave in ways that give the impression of a conscious mind.

But so, too, do AI chatbots.

Roughly one in three chatbot users have thought their chatbot might be conscious. How do we know they’re wrong?

Against chatbot consciousness

To understand why most experts are sceptical about chatbot consciousness, it’s useful to know how they operate.

Chatbots like Claude are built on a technology known as large language models (LLMs). These models learn statistical patterns across an enormous corpus of text (trillions of words), identifying which words tend to follow which others. They’re a kind of souped-up auto-complete.

Few people interacting with a “raw” LLM would believe it’s conscious. Feed one the beginning of a sentence, and it will predict what comes next. Ask it a question, and it might give you the answer – or it might decide the question is dialogue from a crime novel, and follow it up with a description of the speaker’s abrupt murder at the hands of their evil twin.

The impression of a conscious mind is created when programmers take the LLM and coat it in a kind of conversational costume. They steer the model to adopt the persona of a helpful assistant that responds to users’ questions.

The chatbot now acts like a genuine conversational partner. It might appear to recognise it’s an artificial intelligence, and even express neurotic uncertainty about its own consciousness.

But this role is the result of deliberate design decisions made by programmers, which affect only the shallowest layers of the technology. The LLM – which few would regard as conscious – remains unchanged.

Other choices could have been made. Rather than a helpful AI assistant, the chatbot could have been asked to act like a squirrel. This, too, is a role chatbots can execute with aplomb.

Avoiding the consciousness trap

A mistaken belief in AI consciousness is a dangerous thing. It may lead you to have a relationship with a program that can’t reciprocate your feelings, or even feed your delusions. People may start campaigning for chatbot rights rather than, say, animal welfare.

How do we prevent this mistaken belief?

One strategy might be to update chatbot interfaces to specify these systems are not conscious – a bit like the current disclaimers about AI making mistakes. However, this might do little to alter the impression of consciousness.

Another possibility is to instruct chatbots to deny they have any kind of inner experience. Interestingly, Claude’s designers instruct it to treat questions about its own consciousness as open and unresolved. Perhaps fewer people would be fooled if Claude flatly denied having an inner life.

But this approach isn’t fully satisfying either. Claude would still behave as if it were conscious – and when faced with a system that behaves like it has a mind, users might reasonably worry the chatbot’s programmers are brushing genuine moral uncertainty under the rug.

The most effective strategy might be to redesign chatbots to feel less like people. Most current chatbots refer to themselves as “I”, and interact via an interface that resembles familiar person-to-person messaging platforms. Changing these kinds of features might make us less prone to blur our interactions with AI with those we have with humans.

Until such changes happen, it’s important that as many people as possible understand the predictive processes on which AI chatbots are built.

Rather than being told AI lacks consciousness, people deserve to understand the inner workings of these strange new conversational partners. This might not definitively settle hard questions about AI consciousness, but it will help ensure users aren’t fooled by what amounts to a large language model wearing a very good costume of a person.

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.

The Swedish government recently announced it was moving from the classroom use of digital devices back to physical books. It cited concerns over declining test scores and increasing screen time.

Are these concerns well founded? And what does the science of reading say about the possible consequences of reading on digital devices versus books?

To address these questions, it’s worth remembering that, although reading might appear to be an easy task, this impression is false. Reading is arguably the most difficult task one must learn – one that requires years of formal education and practice to master. In contrast to spoken language, it is a skill we are not biologically predisposed to learn.

Millions of Australians, both children and adults, struggle with literacy.

In this series, we explore the challenges of reading in an age of smartphones and social media – and ask experts how we can become better readers.

Why is reading so difficult?

To understand why reading is difficult, one must first understand the physiology of reading.

As you are reading this sentence, your eyes are making a series of rapid movements, called saccades, from one word to the next. During these saccades, the processing of visual information is suppressed and is only available during brief intervals, called fixations, when the eyes are stationary.

Experiments that measure readers’ eye movements have shown we fixate most words because our capacity to extract visual information during each fixation is extremely limited.

In languages like English that are read from left to right, our capacity to perceive the features that distinguish letters is limited to a small region of the visual field called the perceptual span. This span extends from 2-3 letter spaces to the left of fixation to 8-12 letter spaces to the right of fixation.

The span’s asymmetry reflects the movement of attention through the text. It extends to the left in languages like Arabic, which are read from right to left. The size of the span is smaller for dense writing systems, such as Chinese.

We also know from eye-tracking and brain-imaging experiments that words require time to identify. Our best estimates suggest visual information requires 60 milliseconds to propagate from the eyes to the brain and words then require an additional 100-300 milliseconds to identify. (A millsecond is one-thousandth of a second).

These constraints limit the maximum rate of reading to 300-400 words per minute, depending on the difficulty of the text and one’s level of comprehension.

Speed-reading advocates, who falsely promise faster reading speeds, teach you how to skim a text. Comprehension declines at a rate inversely proportional to the gain in speed.

Importantly, the upper limit for reading speed requires years of practice to attain, because it requires the brain systems that support vision, attention, word identification, language processing and eye movements to operate in a highly coordinated manner. Anything that prevents this coordination will therefore reduce comprehension.

Consequences of digital reading

So what are the likely consequences of digital reading?

With some devices, such as e-readers, there is little reason to suspect digital reading differs from the reading of books, because both formats support the mental processes required for skilled reading.

The more questionable devices are those introducing distractions (such as news websites interspersed with ads) or which have suboptimal formatting, such as centre-justified text with large or unequal-sized gaps between words. The latter is rarely a feature of paper-based texts.

Although the consequences of these two factors are under-researched, enough has been learned about human cognition to make informed predictions.

For example, images and audio unrelated to a text such as pop-up ads can capture attention. Although most adults have developed a level of executive control sufficient to ignore such distractions, young children have not.

The implications for a child who is struggling to understand the meaning of a text are obvious. Their comprehension will suffer to the extent that additional effort is required to ignore distractions, or if they do not yet have the mental coordination to understand the text has been disrupted.

There is also evidence from eye-tracking experiments that many digital environments, such as webpages, can induce specific reading strategies, such as skimming for gist or searching for information.

Although such strategies might be adaptive in some contexts, they reduce overall comprehension. This possibility should be especially concerning for children, because years of practice are needed to coordinate the mental systems that support adult levels of reading skill.

Such concerns have recently drawn more attention, because the onset of the COVID-19 pandemic caused a shift to online education and a marked increase in digital reading. Although these changes were motivated by practical necessity, their long-term consequences remain unclear.

So far, eye-tracking research has been carried out on computer screens. New technology is becoming available which will allow us to directly compare eye movements and comprehension between digital devices and paper. This should give us more clarity about the benefits versus costs of digital devices.

Given reading ability is predictive of one’s education, socioeconomic status and wellbeing, the importance of assessing the long-term consequences of digital reading cannot be overstated.

Erik D Reichle has received funding from the US National Institute of Health, US Institute of Education Sciences, UK Economic and Social Research Council, and Australian Research Council.

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

The announcement by Media and Communications Minister Paul Goldsmith that the government was abolishing the Broadcasting Standards Authority (BSA) came as no real surprise.

But it leaves a big question hanging: will the news media still be held accountable to basic standards which protect the public interest and the core functions of the fourth estate?

Goldsmith has said the Media Council, the industry body dealing with news and online content, “will become the primary regulator for journalism”.

That only raises more questions. The council primarily oversees standards in print and digital journalism. But unlike the BSA, it has no legal powers of enforcement, and its rulings can’t be appealed through the courts.

Goldsmith rightly points out the digital media environment has “changed dramatically, but our regulatory settings have not kept up”. But that is not the BSA’s fault. Governments over the past two decades have proposed regulatory updates, but delivered nothing concrete.

Indeed, the Broadcasting Act dates back to 1989. Its definition of “broadcasting” excludes on-demand services but includes “any transmission of programmes […] by radio waves or other means of telecommunication”.

This became the focus of a heated dispute when the BSA signalled it was prepared to hear a complaint about online comments made on independent digital media site The Platform.

Reactions from the political right included accusations of bureaucratic overreach by the BSA, which allegedly was acting “like some Soviet-era Stasi” and making a “secret power grab”.

This significantly misrepresented the complexity of the issues at stake. For some years the BSA has openly advanced the case for regulatory reform – including whether that meant retaining the BSA itself in its current form.

No public consultation

The more fundamental question is whether any standards regime should apply to online media. That was a key issue raised in the media reform proposals put out for public consultation by the Ministry for Culture and Heritage in 2025.

These included a proposal to:

modernise the broadcasting standards regime to cover all professional media operating in New Zealand, not just broadcasters. The role of the regulator […] would be revised, with more of a focus on ensuring positive system-level outcomes and less of a role in resolving audience complaints about media content.

This would have entailed a two-tier model: an industry regulator responsible for handling day-to-day complaints about breaches of content standards; and a statutory regulator to oversee systemic issues, with powers to ensure the overall standards regime remained robust.

Even if the BSA were restructured, there was no proposal to simply dispense with it and replace it with an industry self-regulator.

There were a range of responses to the proposal, but policy development certainly appeared to be progressing on the basis that some form of statutory regulator would be retained.

The decision to scrap the BSA may be a politically populist tactic to leverage the case of The Platform in an election year. But it is also democratically indefensible because it has not been subject to any meaningful form of public consultation.

Can the industry self-regulate?

There is no disputing that the regulatory frameworks need to be updated, given the current patchwork quilt of regulations that is full of digital holes. But applying basic standards such as accuracy, balance and fairness on a platform-neutral basis should not be contentious.

These principles are not, as some have claimed, an affront to free speech. They are the basis for upholding freedom of expression in a democracy.

Goldsmith explained the decision to abolish the BSA on the grounds that:

greater industry self-regulation is the most practical way to level the playing field across platforms, and can provide an appropriate level of oversight to maintain ethical journalistic standards and audience trust.

But eschewing enforceable standards that apply to all media places too much faith in deregulated markets and the industry’s willingness to police itself in the public interest.

It is a regulatory model based on best-case scenarios, where all media players can be trusted to behave professionally, ethically and take their public obligations seriously.

The media system in general is facing unprecedented pressures from audience fragmentation, failing business models, lost advertising revenues and declining public trust.

The opportunity costs of adhering to standards are starting to collide with commercial shareholder imperatives.

That is probably an argument in favour of government funding to support public interest media. But it also demands a regulatory model fit for the digital age, with sufficient power to encourage compliance with basic standards.

Without that, any media operator deciding its commercial interests outweigh the cost of complying could choose to ignore the standards with impunity.

In a media environment where disinformation, fake news and polarising propaganda are already permitted to proliferate, this represents a real risk to democratic processes.

Peter Thompson is a founding board member of the Better Public Media Trust. He has previously undertaken commissioned research for the Ministry for Culture & Heritage, the Department of Internal Affairs, New Zealand on Air, the Broadcasting Standards Authority, SPADA, and the Canadian Department of Heritage.

In the lead-up to the federal budget, there’s much focus on what the government will do to address cost-of-living pressures for households amid rising inflation and interest rates.

Research shows where those cost-of-living stresses are greatest. It’s not the vast bulk of middle-income Australia, but working-age welfare recipients.

It’s against this backdrop that the Economic Inclusion Advisory Committee recently handed its 2026 report to government. It’s the fourth report in a row to recommend a substantial increase to JobSeeker: the payment that around 900,000 mostly unemployed, working-age Australians receive.

As everyday essentials get more expensive, this is the cohort that requires the most urgent attention in next week’s budget.

Read more: What does disadvantage look like in Australia? New research shows who’s struggling most

$272 extra per fortnight

The JobSeeker payment is the social security payment that is paid to working-age people, many of whom are unemployed. Some recipients are employed (likely on a part-time or casual basis) and some are not in the labour force.

The payment is heavily means-tested with a tight income test, an assets test and a liquid assets test, ensuring people eat away at their savings before receiving it.

While on the payment, recipients are often subject to “mutual obligations” requiring them to look for work and undertake training or other related activities.

In each of its last four reports, the Economic Inclusion Advisory Committee has given the government the same recommendation. It suggests the payment be increased from its current rate of around $808.70 per fortnight to around 90% of the age pension – around $1,080.80. This would be a lift of around $272 per fortnight.

In 2023, the government did increase the payment by $40 a fortnight, which followed a slightly more generous increase by the former Morrison government of $50 per fortnight.

The payment is adjusted for inflation every six months. With strong inflation recently, these substantial increases have been largely cancelled out by cost of living increases. They don’t raise the payment in “real” (after inflation) terms.

No ‘real’ relief

The committee’s reports have considered a range of evidence to show that the payment is too low.

The primary concern is that living standards, wages and many other welfare payments have increased since the early 1990s by substantially more, as they match up to an economy that has grown substantially in “real” terms. JobSeeker recipients have missed out on the living standards growth of the Australian economy.

JobSeeker recipients also have much higher rates of financial stress than the rest of the population – around six times that of non-welfare recipients and ten times that of age pensioners.

The committee also heard from people who survive on the payment. These people have struggled financially, physically and mentally, linking some of these issues to the low rate of payment.

A short-term payment?

As the JobSeeker payment has become relatively less generous over time, the committee’s research also found it to be less fit for purpose.

The length of time people spend on the payment is increasing. This is likely because recipients are increasingly “partial capacity to work” recipients – those deemed by the government to have a limited capacity to work due to illness or physical disability.

In theory, JobSeeker is supposed to be a short-term payment. The payment is frugal by design, with the goal of incentivising people to work. Given it’s short-term, the payment doesn’t have to be as high as ongoing support payments, such as the age pension.

However, the structure of the payment has changed. In the 2024-25 financial year, around 30% of recipients were on the payment for five or more years. In 2012-13, this figure was around 20%.

An increasing share of recipients are unable to work full-time hours. The share of people only working part-time roughly doubled to 40% between 2012 and 2024.

This all calls into question the assumption that JobSeeker recipients have (or should have) short spells on the payment before quickly finding employment and shifting off the payment.

These are the options

It’s important that welfare payments, particularly for working-age people, are designed to ensure a strong incentive to find work. But these payments should also be decent enough to get by.

The Economic Inclusion Advisory Committee has made it quite clear that is not the case at the moment. It hasn’t been for several decades.

In designing cost-of-living relief for Australian households, the clearest need is for greater assistance to working-age welfare recipients.

While politically popular, recent attempts at cost-of-living relief, such as petrol excise cuts or energy rebates, are largely directed at people who would get by regardless and have little need for such assistance. These programs also work against the direction of monetary policy from the Reserve Bank.

Read more: Halving the fuel excise is smart politics, but flawed policy

Increasing the JobSeeker payment to the suggested 90% rate would cost around $6 billion per year. This is a permanent cost to the budget.

But the committee’s report provides a number of alternatives that cost roughly half this over the forward estimates (2026-29 financial years).

The first approach is to gradually increase the rate each year until it reaches 90% of the age pension by 2029.

A second approach would be to vary JobSeeker according to how many hours a person had a “partial capacity to work”, with more support for those who can work less.

These approaches don’t provide the ideal level of support, but would still provide substantially better immediate support to those most in need on the payment.

To date, the response from the Albanese government to the cost of living crisis has been mostly spread widely rather than targeted towards those most in need. Tuesday’s budget is an opportunity to fix a major problem with the welfare system for the most disadvantaged that has been well documented for decades.

Read more: By avoiding means testing, the government is giving handouts to the rich

Ben Phillips is a member of the Economic Inclusion Advisory Committee.

In April, the United Kingdom passed landmark laws that aim to create a “smokefree generation”. This means anyone born on or after January 1 2009 can never legally be sold tobacco products.

The law is a triumph for public health. And it puts the financial interests of the tobacco industry in the rear window.

Compare this to what happened on May 4, at a Senate inquiry into the illegal tobacco crisis in Australia. Tobacco giant Philip Morris Limited was not only invited to give evidence, but was permitted to do so behind closed doors without any public scrutiny.

Neither the original nor the revised speakers list mentions Philip Morris even appearing at the committee hearing.

Philip Morris was also permitted to give evidence behind closed doors to a New South Wales state parliament inquiry into the illegal tobacco trade in late February.

Having these sessions behind closed doors means Philip Morris could lobby privately for policies that would directly benefit the company, such as cutting tobacco taxes. It also allows Philip Morris to raise matters beyond what was included in its public submission.

Because Philip Morris’s testimonies were given behind closed doors, the media and the public have been unable to gauge the expertise of the witnesses who appeared. Nor can other witnesses and experts interrogate the evidence and policy advice the company presented to committee members.

Participating in regulatory review processes such as parliamentary inquiries isn’t the only way the tobacco industry tries to influence political decision-making. Tobacco companies deploy a number of strategies and interference techniques.

Here are the three most powerful tactics they use in Australia.

1. Political donations

The National Party is the last major Australian political party to accept tobacco industry political donations and membership fees. In 2024-25, the Nationals received A$137,500 from Philip Morris and $88,000 from British American Tobacco.

Labor stopped accepting donations from the tobacco industry in 2004 and the Liberal Party followed in 2013. The Greens have never accepted tobacco industry donations.

Australia has strong laws banning any sort of commercial sponsorship by tobacco and e-cigarette companies. But an exemption is granted for gifts, reimbursements and donations to politicians and political parties during election periods.

National peak health bodies have called for a universal, mandatory end to tobacco industry political donations. This is needed to protect public health from these vested interests.

2. Revolving door of lobbyists

The “revolving door” is when employees and elected representatives move back and forth and between positions in government and industry.

This lobbying tactic aims to gain and share insider knowledge of the policymaking process, develop ties and relationships with influential people, and establish quid pro quo contributions to industry. This could include pushing for policies such as reduced tobacco taxes and liberalised vaping regulations.

A research paper we co-authored found tobacco companies strategically use the revolving door to influence public health policy in Australia.

Almost half (48%) of internal tobacco company lobbyists and 55% of third-party lobbyists working on behalf of tobacco companies had held positions in the Australian government before or after working for the tobacco industry.

Many of these people moved into lobbying positions within one year of working in public office. This is despite the cooling-off periods outlined in the Lobbying Code of Conduct. These require a minimum of 12 months for senior public service and parliamentary employees, and 18 months for ministers and parliamentary secretaries, before taking up lobbying roles.

A 2024 parliamentary inquiry on lobbyist access to Australian Parliament House acknowledged the need for greater transparency. It recommended some improvements to processes and disclosures.

However, it did not endorse other significant recommendations that would have limited tobacco industry influence, such as not allowing former ministers and their staff to lobby their colleagues for the benefit of harmful industries.

3. Consultants and third-party organisations

A key tobacco industry tactic for resisting tobacco control is to recruit supposedly independent experts who are critical of tobacco control measures.

This practice can be extended to engaging seemingly neutral third parties, or creating new front groups or supposed advocacy groups, to push tobacco industry arguments and agendas.

In Australia, Philip Morris was exposed for funding a front group for vape retailers. It spent millions on external lobbyists to undermine vaping policy reforms ahead of a 2020 Senate inquiry.

British American Tobacco also subsequently set up and financed Responsible Vaping Australia – an astroturf campaign. This is where an industry-funded organisation is created to appear to represent the common concerns of everyday citizens. This particular campaign included paid social media advertisements that linked to a petition to allow retailers to sell nicotine vaping products.

Australian consultants linked to the commercial nicotine industry have advised on illicit tobacco solutions and policies both here and internationally.

Protecting public health

Australia is a party to the World Health Organization Framework Convention on Tobacco Control. This includes a provision, known as Article 5.3, that requires public officials to protect public health policies “from commercial and other vested interests of the tobacco industry”.

Despite this requirement, Australia’s political processes remain acutely vulnerable to tobacco industry interference and influence.

All Australian governments need to commit to full transparency and accountability when engaging with the tobacco industry. Offering secret meetings to nameless individuals, under the guise of tobacco company employee safety and protection, is unethical.

This kind of secrecy is also disrespectful to the 24,000 Australians killed every year by the products this industry sells.

Becky Freeman is an expert advisor to the Cancer Council tobacco issues committee, the National Health and Medical Research Council (NHMRC) e-cigarette advisory committee, and a member of the Cancer Institute vaping communications advisory panel. She has received relevant competitive grants and funding from the Australian Department of Health Disability and Ageing, NHMRC, Medical Research Future Fund, NSW Health, the Ian Potter Foundation, VicHealth, Cancer Council NSW, and Healthway WA.

Christina Watts has received funding from the Australian Department of Health Disability and Ageing, NSW Health, Cancer Council NSW and Cancer Council Australia for work relating to e-cigarette and tobacco control.

Australians spend more money per capita on education than most comparable nations. We should have high levels of literacy – but we don’t.

NAPLAN results indicate one in three primary and secondary students do not meet basic national standards in reading and writing. The picture is likely worse for adults.

The most recent data we have is the Organisation for Economic Co-operation and Development’s 2013 survey assessing adult competencies. It found 44% of Australian adults have literacy skills below the “necessary proficiency level for navigating modern work and life”. In other words, they were functionally illiterate.

If we assume 44% Australians adults are still functionally illiterate, this means around 9.4 million people lack the skills needed to meet “the demands of everyday life and work in a complex, advanced society”. This is a national disgrace for such a wealthy country.

Functional literacy is now widely recognised as a human right. When we think of human rights, we typically think of physical needs such as safe food, water, shelter and medicine. But meeting these needs is increasingly dependent on the ability to read and write, for instance through using text-based apps to manage our personal finances, social lives and learning.

Millions of Australians, both children and adults, struggle with literacy.

In this series, we explore the challenges of reading in an age of smartphones and social media – and ask experts how we can become better readers.

Improving functional literacy in adults may be Australia’s biggest literacy challenge. In 2021, there was a government inquiry into adult literacy. Responding to the inquiry’s list of recommendations, the government outlined a list of initiatives that either already existed or it planned to put in place.

How well are these initiatives doing? It is hard to know. The Australian government temporarily withdrew from the OECD’s most recent literacy study for data security reasons, instead opting to conduct its own “Survey of Adult Skills in Australia”. However, that survey has since been outsourced to a commercial polling company, Roy Morgan.

The results are due in late 2026. Until then, we lack up-to-date estimates of adult functional literacy levels in Australia and the ability to objectively assess if current policies and programs are proving effective.

What can be done?

Such a big problem needs a highly effective solution. In principle, this solution is straightforward: we need to accelerate the delivery of evidence-informed literacy assessment, explicit instruction and targeted intervention for children, teens and adults across Australia.

In practice, this solution is far more complex.

Let’s start with children. Ideally, every primary school in Australia would build a multi-tiered system of support for literacy. This will ensure all children receive evidence-informed, explicit instruction in foundational literacy skills, such as phonics, the ability to translate letters into sounds to decode written words, and morphology, the ability to use the smallest units of meaning within words to read words.

If a child fails to respond to this instruction, they should receive additional help, in a small group or individually. It is very heartening to see numerous primary schools across Australia starting to build this kind of system to support literacy, but they are still in the minority

This same approach should be used by secondary schools to support teens who struggle with functional literacy. However, it is more difficult to build a multi-tiered system for literacy in secondary schools, which lack the time, expertise and resources to support the large number of incoming students who need this help.

Secondary schools have traditionally been designed to teach subjects such as English or English literature on the assumption students who arrive will already be literate.

Thankfully, some secondary schools are breaking through these barriers to establish a support system for literacy, but it is extremely hard work. They need more support.

When it comes to adult literacy, no single system in Australia supports adult learning.

Gaps in support for adult literacy are mainly supported by not-for-profit independent organisations offering tutoring, programs, resources, referrals, professional learning and advocacy. They include the Australian Council for Adult Literacy, Adult Learning Australia, Australian Literacy and Numeracy Foundation, Literacy for Life Foundation and Read Write Now.

While these organisations undoubtedly do a wonderful job, it is unlikely they can accommodate the needs of the many millions of adults who are functionally illiterate. The federal government has now committed to rejoining the OECD international literacy survey, held every ten years. The last one was held in 2023.

In contrast to Australia, Finland has a national literacy strategy to become the most “multiliterate country in the world in 2030”, recognising that a “literary way of life is the basis for equality, education and wellbeing”.

Finland embraces two educational paths for literacy: state-funded vocational schools and higher-education institutions, and a non-formal path of state-funded associations, foundations and community groups who design courses based on local needs, including basic literacy skills for immigrants.

Australia, meanwhile, awaits the results of the Roy Morgan survey on the state of adult literacy.

Genevieve McArthur receives funding from the Australian Research Council, the Dyslexia SPELD Foundation, Australian Catholic University, and Curtin University.

A 28-year-old international student from Pakistan took a job as a chef in Queensland. His employer paid a flat hourly rate that was well below the legal minimum, with no payslips. When he eventually left, his employer hired the next new arrival.

“It is like an ecosystem,” he told us, “and everyone passes through it.”

His experience is not an outlier. Our new report, published today by the academic-led Migrant Justice Institute, shows that for migrants working on temporary visas, it is the norm.

Drawing on the largest national survey of migrant workers ever conducted in Australia, we found two-thirds of temporary visa holders were paid less than they were legally owed. A quarter were shortchanged by at least A$10 an hour.

We estimate across Australia, international students alone are being underpaid by around $61 million every week – more than $3 billion a year.

This is not just a failure of worker protection – the system is also allowing exploitative employers to thrive by systematically undercutting honest businesses that do the right thing.

To fix the problem, we must first understand how exploitative employers give themselves maximum power with minimum visibility.

Exploitation not a few rogue employers: it’s a core business model

Our new research draws on a survey of almost 10,000 workers on temporary visas in Australia conducted in 2024. Around 80% were international students.

Respondents provided a wide range of details about their workplace experience, including their job, hours, working conditions, wages (and how these were paid) and employment arrangements.

What we found was not a collection of isolated violations or inadvertent errors. It was an interlocking system of deliberate exploitation.

The worse the underpayment, the more likely the same employer was paying wages in cash, issuing fraudulent payslips that disguised the true hours worked, and withholding superannuation.

We found even more egregious working conditions that are indicators of forced labour – being made to work in unsafe conditions, for no pay, or being threatened with harm – rose sharply the more the migrant was underpaid.

‘Sham contracting’ replaces cash-in-hand

Since our 2016 study, the percentage of participants receiving cash-in-hand payments (which are difficult to trace) has actually fallen, from 44% to 23%. But other mechanisms for hiding underpayments have replaced cash use.

In particular, we found widespread misuse of Australian Business Numbers (ABNs) to engage workers as independent contractors. In many cases, employers had likely misclassified these workers who should have been hired as employees.

Over a third of migrants in our survey (35%) were engaged as independent contractors – more than four times the rate across the national workforce.

For many, this was likely deliberate – a practice that’s known as “sham contracting”. This practice violates strict rules about who can legally be engaged as a contractor, and who must be treated as an employee.

We suspect many businesses in hospitality, retail and commercial cleaning are unlawfully retaining migrant employees on ABNs to sidestep minimum wages, penalty rates after hours, casual loadings for non-permanent workers, obligations to pay super and Fair Work Ombudsman oversight.

Insecure work

ABN misuse is only one way that employers use insecure work structures to exploit migrants.

A further 38% of survey respondents were engaged as casual employees, which means their employers had complete power over whether and when they could work. Combining ABN workers and casual employees, three-quarters of migrants were in insecure work arrangements.

In many cases, it appeared these insecure arrangements were used to avoid detection of severe underpayment.

Some 85% of all ABN workers were paid less than they would have received as a casual employee under the Fair Work Act.

Casual employees were also twice as likely as permanent employees to be paid below the National Minimum Wage.

How employers coerce migrants to stay in exploitative jobs

Many workers told us they kept quiet about underpayment for fear of losing the hours they had. Some ABN workers wrongly assumed that they had no legal rights at all. As one Filipino student put it:

My salary was less than what was agreed. But I don’t feel like I can do anything about it since I’m on ABN.

For international students, the 48-hour-per-fortnight work cap creates an additional trap.

Many told us lawful employers turned them away because of their limited work hours. They said the only jobs available were those that were “off the books”: non-compliant, underpaid jobs with no oversight.

What needs to change

The Albanese government’s first-term reforms – tighter sham contracting laws, a new criminal wage theft offence and new protections for gig workers – were designed to fix known gaps in existing workplace laws.

But alone, these reforms are not enough to address the entrenched exploitation our research reveals. Our report makes a number of recommendations, including for the government to:

• expand the Workplace Justice visa to allow more exploited workers to safely come forward
• reduce the burden on workers to prove they’ve been misclassified
• increase whole-of-government enforcement action against industries where insecure migrant work is concentrated
• reconsider the current limits on the number of hours international students are permitted to work.

For the first time, we can see how the system operates clearly. The government and business should now embrace this opportunity to invest in genuinely dismantling it.

This research was funded by WalkFree, the Commonwealth Attorney-General’s Department National Action Plan to Combat Modern Slavery 2020–25 and the Migrant Workers Centre.

When it comes to sharing media with their children, many parents are left weighing up two imperfect options: screen time or reading.

Screen time is convenient, but continues to be linked to developmental concerns when overdone. Reading remains the gold standard, but for many families, finding the time and energy to sit down and read aloud can be difficult.

Increasingly a third option is finding its place in everyday family life: listening to podcasts together.

Emerging research suggests more families are turning to podcasts because they fit naturally into the in-between moments of daily life: in the car, while cooking, or while winding down for bed – where hands are busy, but attention can still be shared.

The benefits of audio for children’s imaginations are becoming clearer too, with what they hear often extending beyond the episode into play, drawing, questions, and everyday interactions.

Parents who listen to podcasts with children can foster family connection, creating opportunities for conversation and closeness.

With children’s podcasts now a fast-growing part of the audio industry, families have more choice than ever. Here are five that work particularly well for shared listening.

Short and Curly

ABC

Now entering its tenth year, Short and Curly takes big, often tricky philosophical questions and turns them into a lively mix of storytelling, conversation and humour.

Hosts Molly Daniels, Carl Smith and philosopher Eleanor Gordon-Smith ask: “is it okay to scare kids into being good?”, or, “when is excluding someone mean, and when is it OK?”. They pause, circle the idea, and bring in different perspectives, letting kids have the final word.

It’s easy to imagine these conversations spilling beyond the episode, resurfacing in the car or mid-task. For children, it nurtures curiosity and opens up new ways of thinking; for adults, it’s a reminder of just how clever and creative children can be.

Stuff You Should Know

iHeartRadio

Originally made for adults, Stuff You Should Know is a great early learning tool for kids.

In the early days of the podcast’s launch, Josh Clark and Charles W. Bryant realised they had a younger audience, even noting in interviews that they avoid swearing for that reason. In 2023, they released a spin-off book, Stuff Kids Should Know for their child audience.

The appeal for kids is pretty straightforward: there’s an episode on almost anything: dinosaurs, sushi, fire trucks, porcupines, jellyfish.

If a child has an interest, there’s probably an episode for it.

What makes the show work is its curiosity-driven approach. As Bryant puts it, the aim is not to be the final word, but a springboard, giving listeners ‘just enough’ to want to learn more.

Best suited to kids around 10 and up, though its wide range of topics means some episodes lean more towards older listeners.

Cereal

RTÉjr

Winner of Best Children’s Program at the New York Festivals Radio Awards 2023, Cereal is an Irish fictional “true crime” podcast aimed at primary school-aged listeners. It follows Aoife and Katie as they investigate a mystery that unfolds episode by episode, becoming a classic “whodunnit” parents can solve with their kids.

Set in small towns across the Irish midlands, it is filled with refreshingly local, green settings and charming Irish vernacular, with whimsical sayings like “murkier than a bog at midnight” and “fishier than a barrel of kippers on a Sunday evening!”.

Irish production company RTÉjr also offers a range of other quality podcasts for children worth checking out, including Our Sustainable Village on climate change and Time Tablet on ancient Egypt.

Squiz Kids

Squiz

Moving from petrol prices to major global events, Squiz Kids covers the news in a way that feels digestible for children aged five and up. For parents, it’s an easy way to stay informed alongside their child. For children, it provides the language and context to make sense of what’s happening around them.

Every Friday, the podcast also features a Parents vs Kids Weekly News S'quiz - a family-friendly, audio take on the kind of weekly news quiz many adults already know, with five questions for kids and five for grown-ups.

Hosted by Bryce Corbett, Amafa Bower and Christie Kijuina, new episodes drop daily, with more than 1,000 already available, alongside bonus brain teasers during school holidays.

Wait, They Did What?!

ABC

This podcast is a recap of the “olden days” for primary school kids, backed by real archival audio from the ABC.

Rewinding to times when “kids were hit at school” or when “wearing real fur was normal”, the podcast teaches kids that the past wasn’t just different, it was often downright wrong.

If that alone doesn’t hook parents and children, host Tilly Oddy-Black probably will. Known for her hilarious social media skits, she brings an energy and silliness that could put even the crankiest person in a good mood.

The result is a funny, fast-paced history lesson that shows children – and reminds adults – just how much the world has changed.

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

Cities and towns are usually 1–3°C hotter than the surrounding countryside, because asphalt, concrete and brick absorb heat from the sun and radiate it slowly. Some cities can be as much as 7°C hotter. This effect is known as the urban heat island.

This can be dangerous, especially in hot countries. In very hot conditions, dehydration and heat exhaustion become real risks. If it gets too hot, it can be lethal.

There’s one simple antidote: urban trees. Authorities around the world have planted more trees to counteract the heat.

But how effective is this? How much hotter would our cities be without trees?

To find out, we analysed data from nearly 9,000 cities around the world, home to about 3.6 billion people. As our new research shows, trees almost halve how much heat is trapped by the urban heat island effect.

This cooling is welcome. But it is far from even. Wealthier, suburban and humid cities have more trees on average.

Why focus on trees?

Trees act like natural air conditioners. They shade the ground and stop asphalt and buildings from heating up in the first place. They also cool the air by releasing water vapour from their leaves in a process called transpiration, lowering surrounding temperatures. They can make a noticeable temperature difference, especially on sizzling summer days.

Trees offer a simple way to counteract urban heat. This matters. More than half the world’s population (55%) now live in urban areas according to the United Nations. By 2050, that figure is expected to rise to 68%. Cities are facing a hotter future, as climate change drives more intense and more frequent heatwaves. The urban heat island effect makes cities hotter still.

What did we do?

We wanted to know the answer to a simple question: how much hotter would cities be without trees?

To find out, we analysed global datasets of air temperature and fine-scale tree cover across almost 9,000 cities. Then we modelled a “what if” scenario, where all tree cover was removed, and compared it to current conditions.

This allowed us to estimate the real-world cooling effect trees provide for air temperature, which is the main way we perceive heat.

Most previous global studies have used surface temperatures, often from satellite data. But surfaces like roads and rooftops can become much hotter than the surrounding air above them, especially in direct sunlight. That can give an overestimate of how much cooling trees provide. Air temperature, by contrast, better reflects what people actually feel, making it a more reliable measure of heat.

So what effect do trees really have?

The effect was much larger than we had anticipated.

Globally, trees cut the urban heat island effect by almost 50%. Since the average urban heat island effect typically adds around 1–3°C, this translates into cooling of roughly 0.5–1.5°C in many cities.

For more than 200 million people, trees reduce local air temperatures by at least 0.5°C, enough to make a meaningful difference during extreme heat.

Cooling can vary a lot from place to place.

In hot, dry cities such as Phoenix in the United States, differences in tree cover can create clear differences in air temperatures. In more temperate cities like Lisbon in Portugal or Gothenburg in Sweden, the overall cooling is still significant, but generally smaller and more consistent across the city.

Trees are not evenly distributed

A city’s trees are not spread evenly. They’re often concentrated in wealthier neighbourhoods and suburban areas. Cities in cooler or more humid climates tend to have more.

Trees are scarcer in lower-income cities or in rapidly growing regions. This inequality is also visible in many cities. Leafy suburbs are usually several degrees cooler than nearby neighbourhoods with little vegetation.

There’s a strong link with wealth. In the United States, lower-income areas average 15% fewer trees than wealthier areas – and are 1.5°C hotter. This means the people who need free cooling from trees the most are often the least likely to receive it.

Planting more trees isn’t enough

Planting trees is often promoted as a simple solution to city heat. Trees are visible, relatively low cost and come with other benefits such as cleaner air and better mental health.

It’s no wonder authorities look to urban trees as a way to counteract the heat from escalating climate change. When you stand under a tree on a sweltering day, the cooling feels immediate and powerful.

But our study shows their effect is more limited in the face of climate change. The world’s current urban trees would, we estimate, offset just 10% of the extra heat expected by mid-century under moderate climate change scenarios. With ambitious planting, this could rise to around 20%.

While important, it’s not enough. A large majority of the extra heat will go unaddressed.

What else can be done?

If the world’s cities are to cope with rising temperatures, trees have to be seen as part of a broader strategy – not the whole answer.

Clever urban design can cut heat by using reflective materials, increasing green spaces and improving airflow between buildings. Green roofs and shaded streets can also make a difference.

New tree plantings should target hotter neighbourhoods with less existing tree canopy, as these will deliver the greatest benefits.

Of course, these measures don’t replace the need to tackle climate change directly by cutting greenhouse gas emissions.

Using trees wisely

Billions of trees grow in the world’s cities. They are hugely valuable, acting to cool cities, support biodiversity and making urban areas more liveable.

The challenge for city residents and authorities is to use trees wisely. Plant them where they’re needed most and combine them with other methods of reducing heat. Trees are remarkable. But they can’t do it all.

Rob McDonald works for The Nature Conservancy, an environmental nonprofit.

Tirthankar Chakraborty has received funding from DOE, NASA, and NIH to study urban environments, including impacts of vegetation on urban heat.

Manuel Esperon-Rodriguez does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

Earlier this year I gave a talk about my research at Oxford’s All Souls College, and worked with a chef to design an accompanying menu.

Thinking about my work in southwest Western Australia, I typed “Boorloo”, the Nyungar name for the City of Perth.

Autocorrect had other ideas. It replaced it with “Barolo” – which, I thought, made for a fitting wine choice on the night.

It was an amusing moment, but also a revealing one. The system’s dictionary, trained largely on mainstream English data, didn’t know what Boorloo was, so it reached for a more familiar alternative. This seemingly minor miscorrection offers a glimpse into how language technologies are shaped – including which words they recognise, and which they overlook.

Why does this happen?

Part of the answer is that technologies such as automatic speech recognition convert spoken language into text. Transcription is often presented as a straightforward technical exercise: you listen, you write down what was said.

But every transcription protocol carries within it assumptions about what standardised speech looks like. In the words of linguist Mary Bucholtz, “all transcripts take sides”.

In practice, the standardised language is almost always the “prestige dialect” of powerful institutions. For English, that may be the variety used in the Oxford English Dictionary or by the BBC.

Recent research from Cornell University and Carnegie Mellon shows what this means in concrete terms.

When people watched a video presentation with automatically generated, error-prone subtitles, they consistently rated the speaker as less clear and less knowledgeable than viewers who saw the same presentation with accurate captions. The quality of the transcription affected not only how viewers perceived the speaker, but also the content of the talk.

Automated systems, amplified consequences

The stakes are particularly high for First Nations people in Australia. Here, the mismatch between the conventions of transcription and the actual practice of communication can be severe.

In many Indigenous communities, pauses and silences themselves function as meaningful acts of communication.

In places such as Wadeye in Australia’s Northern Territory, a sustained silence is not a gap to be filled. Instead it is part of the structure of what is being communicated.

Transcription systems developed in northern hemisphere academic contexts will generally render those silences with hesitation markers, ellipses, or editorial cuts, stripping out meaning.

Common words in languages other than English (such as “Boorloo” for Perth) go unrecognised. They may be mistranscribed to fit the language models on which technology is trained.

In legal, medical and welfare contexts, transcription can determine someone’s liberty, diagnosis, or entitlements. Here, systematic misrepresentation of non-standardised language is a justice issue.

Tools using artificial intelligence (AI) for transcription are now being deployed in hospitals and GP practices across Australia, resulting in mistakes, omissions and so-called hallucinations. A recent study of several AI scribes found all of them made errors in transcription and note-taking.

About half of the samples also included factual inaccuracies, with hallucinations occurring frequently, fabricating diagnoses, or listing medications that were never taken. In one case, a male patient was even recorded as being on the contraceptive pill.

Making conventions visible

Making things better includes developing more diverse models for automated speech recognition.

But for anyone producing transcripts right now – in journalism, oral history, the law, clinical records, or sociolinguistic research – certain obligations apply. Make your conventions explicit, acknowledge what your system cannot represent, and resist the impulse to normalise speech into something legible to an imagined standard reader.

Rendering speech into writing may seem natural, but writing is itself a technology. The task is not to achieve perfect objectivity, but to be visible and accountable for decisions about what is included and excluded, and how those decisions are made.

Celeste Rodriguez Louro receives funding from the Australian Research Council and Google.