If artificial intelligence (AI) systems shape decisions that affect people’s lives, they should do so fairly. This should be a given considering that potential applications for AI include automated hiring systems, as well as tools used in education, finance and criminal justice.

But ensuring the fairness of AI systems is far more complex than it might sound. Despite years of research, there is still no consensus on what fairness means, how it should be measured, or whether it can ever be fully achieved.

Fairness inherently depends on context. What counts as fair in one domain may be inappropriate or even harmful in another. In criminal justice, fairness may prioritise avoiding disproportionate harm to particular communities. In education, it may focus on equal opportunity and long-term outcomes.

In finance, it often involves balancing access to credit with risk assessment. Because AI systems must be formalised mathematically, researchers translate fairness into technical definitions expressed through metrics that specify how outcomes should be distributed across groups.

These metrics are useful tools, but they are not neutral. Each encodes assumptions about which differences matter and which trade-offs are acceptable.

Problems with the data

A deeper issue lies in the data itself. AI systems learn from historical datasets that reflect past decisions, institutional practices, and social inequalities. When a model is trained to replicate observed outcomes, such as hiring decisions or loan and mortgage approvals, it may reproduce existing injustices under the appearance of objectivity.

Optimising for one notion of fairness often means violating another. This tension is evident in automated loan approval systems. An algorithm may be designed so that applicants with the same predicted probability of default are treated similarly across demographic groups.

Yet one group may still be more likely to be incorrectly denied credit, while another may be more likely to receive loans they later struggle to repay. Fairness in predictive accuracy can therefore conflict with fairness in how financial risk and opportunity are distributed.

These differences often reflect structural inequalities embedded in the data the model is trained on. Groups that have historically faced barriers to credit, due to factors such as discrimination or exclusion from financial systems, may have thinner credit histories or lower recorded incomes.

As a result, models can treat socioeconomic disadvantage as a signal of higher risk, even when it does not reflect an individual’s actual ability to repay.

The same pattern emerges in hiring. If a company historically promoted fewer women into senior roles, a system trained to predict “successful” candidates may learn patterns that favour characteristics more common among men, even if gender is not explicitly included as an input. In both cases, the model does not invent bias, it inherits it.

A fundamental question is whether AI systems mirror the world as it was, or attempt to correct for known injustices.

The idea of fairness is further complicated by how it is assessed. Many assessments examine a single protected attribute, such as gender or race, in isolation. While common, this approach can obscure how discrimination operates in practice.

An automated hiring system might appear fair when comparing men and women overall, and fair when comparing ethnic groups overall, yet it might also consistently disadvantage older women from minority backgrounds.

Complex evaluation

People are defined by several characteristics that intersect, including age, ethnicity, disability, and socioeconomic background. Because these intersectional subgroups are often small and underrepresented in data, the harms they face may remain invisible in standard evaluations.

This invisibility has a direct technical consequence. When a subgroup is small, the model encounters too few examples to learn reliable patterns for that group and instead applies generalisations drawn from the broader categories it has seen more of, which may not reflect that group’s actual characteristics or circumstances.

Errors and biases affecting small subgroups are also less likely to surface in standard performance metrics, which aggregate results across all users and can therefore mask poor outcomes for minorities within minorities. Which means that those most at risk are therefore often the least visible.

These challenges suggest that fairness in AI cannot be reduced to better metrics or more sophisticated algorithms. Fairness is shaped by institutional context, historical legacies, and power relations.

Decisions about what data to collect, which objectives to optimise, and how systems are deployed are influenced by social and organisational factors. Technical fixes are necessary but insufficient. Meaningful approaches must engage with the broader context in which AI systems operate.

This includes involving interested parties beyond engineers and data scientists. People affected by AI systems, often members of marginalised communities, possess contextual knowledge about risks and harms that may not be visible from a purely technical perspective.

Participatory approaches, in which affected groups contribute to the design and governance of AI systems, acknowledge that fairness cannot be defined without considering those who bear the consequences of automated decisions.

Even when interventions appear successful, they may not remain so. Societies change, demographics shift and language evolves. A system that performs acceptably today may produce unfair outcomes tomorrow. In particular, recent advances in large language models, the technology underlying many widely used AI tools, add further complexity.

Unlike traditional systems that make specific predictions, these models generate language based on vast collections of historical text. Such datasets inevitably contain stereotypes and imbalances.

Fairness is therefore not a one-time achievement but an ongoing responsibility requiring monitoring, accountability, and a willingness to revise or withdraw systems when harms emerge.

Together, these challenges suggest that fairness in AI is not a purely technical problem awaiting a finite solution. It is a moving target shaped by social values and historical context.

Rather than asking whether an AI system is fair in the abstract, a more productive question may be: fair according to whom, under what conditions, and with what forms of accountability? How we answer that question will shape not only the systems we build, but the kind of society they help to create.

Michael Mayowa Farayola receives funding from Taighde Éireann Research Ireland grants 13/RC/2094_P2 (Lero) and 13/RC/2106_P2 (ADAPT) and is co-funded under the European Regional Development Fund (ERDF).

We asked ten literary experts to recommend the climate poem that has spoken to them most powerfully. Their answers span over 200 years and a range of emotions from sorrow, to anger, fear and hope.

This article is part of Climate Storytelling, a series exploring how arts and science can join forces to spark understanding, hope and action.

1. Death of a Field by Paula Meehan (2005)

Published in the wake of the 2008 financial crisis, Paula Meehan’s Death of a Field critiqued the environmental impact of the Celtic Tiger economy in Ireland.

The poem anticipates the destruction of the titular field by property developers with little regard for native ecologies: “The end of the field as we know it is the start of the estate.”

The global effects of the climate crisis are seen from a uniquely local perspective as the displacement of Irish wildlife mirrors the effect of colonial violence. “Some architect’s screen” is simply the latest iteration of imperial technologies that seek to plunder Irish landscapes. The poem gains further strength by refusing to replicate a hierarchical relationship to nature by preserving its many mysteries:

Who can know the yearning of yarrow

Or the plight of the scarlet pimpernel

Whose true colour is orange?

Jack Reid is a PhD Candidate in Irish literature

2. Darkness by Lord Byron (1816)

Darkness imagines the fallout of a volcanic eruption that has destroyed the Earth. The “dream” that the poem mentions was inspired by genuine weather conditions during the “year without a summer” in 1816, caused by the eruption of Mount Tambora in Indonesia the previous year.

Sulphur in the atmosphere caused darkness and low temperatures across Europe. In Lake Geneva, Lord Byron experienced the infamous “haunted summer” of darkness.

Byron’s depiction of climate catastrophe is bleak, with words like “crackling”, “blazing” and “consum’d” bearing resemblance to contemporary reports of wildfires caused by climate change. After a famine, all elements of Byron’s Earth, from the clouds to the tide, eventually cease to exist: “Seasonless, herbless, treeless, manless, lifeless– / A lump of death – a chaos of hard clay.” Read as a portent of the Anthropocene, Byron’s poem urges readers to seriously consider the future of mankind.

Katie MacLean is a PhD candidate in English Literature

3. Mont Blanc by Percy Bysshe Shelley (1817)

Byron’s close friend Percy Bysshe Shelley was also inspired by the “year without a summer”. He witnessed temperatures drop, volcanic ash hanging heavy in the air and crops failing. While his wife Mary used the gloomy climatic event to inform her novel Frankenstein (1818), Shelley channelled them into his poem Mont Blanc.

In his ode, Shelley describes a timeless “wall impregnable of beaming ice”. By drawing on his scientific reading, he then explains his fears regarding global cooling and the possibility of vast glaciers eventually covering the alpine valleys.

He imagines “the dwelling-place / Of insects, beasts, and birds” being obliterated and mankind forced to flee. While Shelley saw this process as “destin’d” and inevitable, it is clear that Mont Blanc is a poem with catastrophic climate change at its heart. In 2026, it is difficult to read in any other way.

Amy Wilcockson is a research fellow in Romantic literature

4. Characteristics of Life by Camille T. Dungy (2012)

There’s something gloriously elastic about invertebrates: the spinelessness of a worm, the pulsing of the jellyfish, the curling of an octopus. Spiders, snails and bees, too, with their exoskeletons on display, invite us to see things “inside-out”.

These are the thoughts I have when I read Characteristics of Life by Camille T. Dungy, which opens with a snippet from a BBC news report claiming that “a fifth of animals without backbones could be at risk of extinction”. What would a world be without the “underneathedness” of the snail beneath its shell beneath the terracotta pot in the garden? Or “the impossible hope of the firefly” whose adult lives span only a handful of human weeks?

Dungy speaks from a “time before spinelessness was frowned upon”, and from a world where to dismiss a being as “mindless” (jellyfish have no brains) or even “wordless” would be “missing the point” entirely. As I think of these creatures that dwell beyond our usual line of vision – flying, crawling, tunnelling and swimming – I find my perspective on our beautiful world turning and shifting.

Janine Bradbury is a poet and a senior lecturer in contemporary writing and culture

5. Prayer at Seventy by Vicki Feaver (2019)

One of my favourite poems about climate change is Vicki Feaver’s Prayer at Seventy from her 2019 collection I Want! I Want!.

The speaker’s request of passing her “last years with less anxiety” appears to be denied by a god who first responds by changing her into “a tiny spider / launching into the unknown / on a thread of gossamer” and who, when she begs to “be a bigger / fiercer creature”, turns her into “a polar bear / leaping between / melting ice floes”.

Both images present creatures who are in precarious positions, their futures uncertain, reflecting the state of a person contemplating the unknowns of old age and death. But the poem moves beyond the personal. The reference to the melting ice floes is not solely metaphorical: it reminds us that the planet itself is in danger and every living thing is therefore vulnerable – and will be increasingly so.

Julie Gardner is a PhD candidate in literature

Read more: How poetry can sustain us through illness, bereavement and change

6. Walrus by Jessica Traynor (2022)

Walrus, from Jessica Traynor’s 2022 collection Pit Lullabies expresses the quiet anxiety a mother has for her child in the world of climate breakdown.

While stripping wallpaper from the box room of her house, the poet discovers a mural of the Walrus and the Carpenter from Alice’s Adventures in Wonderland. Traynor takes part of Lewis Carroll’s poem about the Walrus and the Carpenter walking along the beach, eating the vulnerable oysters, and weaves it into her own poem.

Carroll’s absurd verse includes what, at that time no doubt, seemed like an impossible image of a “boiling hot” sea. In the 21st century, this is no longer an absurdity, as Traynor knows. She makes a connection with Carroll’s poem, imploring her child:

Sleep as the sun rises and ice melts

and for want of the freeze a walrus

pushes further up a cliff-face.

It’s a complex poem that reimagines a key work of children’s literature, connecting it with the reality of the changing world. All the while the mother keeps her fears at bay for the sake of her child, “brows[ing] washing machines” with a “ball of tears” in her throat.

Ellen Howley is an assistant professor of English

7. Ocean Forest, co-created by the We Are the Possible programme

Ocean Forest is woven out of words, research, ideas and stories shared by scientists, educators, health professionals, youth leaders, writers and artists. They took part in creative writing workshops to co-create the anthology Planet Forest – 12 Poems for 12 Days for the UN Climate Conference in Brazil in 2025.

In the shallows, alert to change,

the minuscule, overlooked creatures

weave between seagrass, and weed –

live their shortened lives.

When ships pass overhead, when sands shift,

fish navigate swell, migrate beyond

where coral’s been bleached, through schools

of silenced whales and barely rooted mangroves

struggling to thrive in darkening water.

Deeper down,

pressure builds, species exist, unaware,

undisturbed. As heat and waves rise there’s hope

the unfound, the unnamed, the unpolluted

in the remotest ocean forests will survive.

Through uniting disciplines and voices the poem takes unexpected shifts. It demonstrates that climate change affects and erodes the habitats that lie beneath the surface and that urgent action is needed to protect disappearing species.

Yet, there is also a glimmer of hope – that in the deepest, darkest parts of the ocean, where temperatures are near freezing and there are bone-crushing pressures, maybe there are creatures that will survive human interference and pollution.

Sally Flint is a lecturer in creative writing and programme lead on the We Are the Possible programme

8. Di Baladna (Our Land) by Emi Mahmoud (2021)

Emtithal “Emi” Mahmoud is a Sudanese poet and activist, who has won multiple awards for her slam poetry performances. Mahmoud performed Di Baladna at the United Nations Climate Change Conference in 2021.

Poetry – especially spoken word – helps people connect emotionally with the human side of climate-driven displacement, a topic that’s often explained only through technical language. The language of emissions targets, temperature thresholds, or policy frameworks can distance people emotionally from its consequences. Yet poetry can cut through this abstraction.

Mahmoud’s performance gave voice to those forced from their homes by environmental collapse, reminding listeners that climate change is not only an environmental crisis but a deeply human one, with profound effects on individuals, families and communities.

By merging vivid natural imagery with the rhythms of displacement and lived testimony, the poem urges listeners to replace passive awareness with empathy. Mahmoud implores us to feel the loss, fear and resilience of displaced communities, looking beyond news headlines and images of victimisation. Engaging with such work helps transform climate refugees from statistics into people.

Clodagh Philippa Guerin is a PhD candidate in refugee world literature

9. Flowers by Jay Bernard (2019)

At first glance, Jay Bernard’s Flowers is circular poem (one that begins and ends in the same place) but you soon realise that the circle isn’t going to complete. It opens:

Will anybody speak of this

the way the flowers do,

the way the common speaks

of the fearless dying leaves?

And closes:

Will anybody speak of this

the fire we beheld

the garlands at the gate

the way the flowers do?

And the answer seems to be, no: no one will speak of these things – the “coming cold” and the “quiet” it will bring – only the things themselves as they die. With the songs Where Have All the Flowers Gone? by Pete Seeger and Blowin’ in the Wind by Bob Dylan in its DNA, Flowers has the eternal power of a folk-lyric – prophetic and unignorable.

Kate McLoughlin is a professor of English literature

10. Place by W.S. Merwin (1987)

Climate change poetry – should it be a thing? How do poets avoid the oracular pomp it threatens? Browsing my small library I’m shocked anew to realise most poets lived and died blissfully innocent of our condition.

OK, what about the late John Burnside’s lyric Weather Report (“this is the weather, today / and the weather to come”). It poignantly extrapolates from a sodden summer to his sons’ futures: “a life they never bargained for / and cannot alter”. Heartbreaking. Or the odd dread of spring in Fiona Benson’s Almond Blossom, a season characterised as Earth’s, “slow incline … inch by ruined inch”. Ditto.

But then I reach back to the great American poet W.S. Merwin’s short prayer Place to find that grace-note of hope which surely needs to thread through all poems, whether they speak of climate change, mortality or love: “On the last day of the world / I would want to plant a tree.” Me too.

Steve Waters is a playwright and professor of scriptwriting at the University of East Anglia

This article features references to books that have been included for editorial reasons, and may contain links to bookshop.org. If you click on one of the links and go on to buy something, The Conversation UK may earn a commission.

Amy Wilcockson receives funding from Modern Humanities Research Association as Research Fellow for the Percy Bysshe Shelley Letters project.

Steve Waters receives funding from AHRC

Clodagh Philippa Guerin, Ellen Howley, Jack Reid, Janine Bradbury, Julie Meril Gardner, Kate McLoughlin, Katie MacLean, and Sally Flint 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.

Twenty years after the first instalment catapulted Anne Hathaway and Emily Blunt onto Hollywood’s A-List, The Devil Wears Prada is back with a second incarnation. The sequel reunites the pair with Meryl Streep and Stanley Tucci for a fun, frothy – but not very devilish – time.

Set at Runway, a thinly veiled fictional version of Vogue magazine, much has changed in the world of journalism since the first film was released in 2006.

Anne Hathaway’s Andy Sachs has spent the intervening years becoming a “Serious Journalist”, with awards galore under her belt. In 2026’s precarious media landscape, though, her job is wiped out. She, somewhat miraculously, finds herself back at Runway as features editor, no longer a harried underling.

Delightfully, the gang is back together for part 2. The Devil Wears Prada’s mastery was always its actors, and the returning main cast are in fine form here. Andy (Hathaway) now has an assured confidence that was just budding in the first film.

The growth in her character is believable and realistic, and as an actor, Hathaway is edging towards greatness, one teary-eyed smile at a time. Andy’s elevated position at Runway allows the dynamic between her and her icy boss, Miranda Priestley (Meryl Streep), to shift.

Miranda de-fanged

Fun is poked at Miranda’s behaviour, which is now subject to HR rules and regulations. Where once she struck fear into the hearts of all she encountered, delivering caustic lines in a low sardonic murmur, Streep’s performance, while fuller and more rounded, de-fangs Miranda.

With disappointingly fewer barbs, she is less “devil”, delivering a more complex portrait of a successful woman struggling to keep a dying industry afloat. Much of the villainy is handed instead to Emily (Emily Blunt). All eye rolls and sharp edges, Blunt has a ball reprising the role that made her a star.

She is given more screen time in this instalment, with a love interest and a life outside of work. She is magnetic in every frame she inhabits, bringing comedy and deliciously over-the-top cattiness.

Stanley Tucci’s Nigel, a relic of the bygone days of print fashion journalism, radiates a warmth that grounds the film. His endless patience with the nonsensical behaviour of those around him, delivered with Tucci’s characteristic panache, steadies the ship when all threatens to spiral into parody.

In 2026, the romantic comedy is a lesser spotted animal in Hollywood compared to when the first film was released. This sequel recalls familiar tropes of the early noughties rom-com: pop music blaring over street scenes of characters speaking on phones, quick cuts between fashion shows and urban life, big cities rendered in gloriously lit night scenes.

The “rom” part of rom-com, though, could have been left in the past for this sequel. Patrick Brammall is criminally underused as Peter, a love interest for Andy. Their dalliance adds little to her character or the story, and never meaningfully develops or resolves.

Journalism SOS

Story-wise, it feels as though the film-makers wanted to comment on the state of journalism. In today’s world awash with algorithms, misinformation and the relentless churn of online content, there was certainly potential to mine, but these themes are mentioned and then glossed over.

This would be forgivable, given the sugary tone of the film, but consequently the drama becomes a little convoluted and at times gets in the way of the relationship dynamics, which is really why we are all in the cinema in the first place. Minor characters played by B.J. Novak, Kenneth Branagh, Lucy Liu and Justin Theroux often lean too far into caricature and disrupt the tone of the film. Their inclusion is another unnecessary dilution of the core four’s chemistry.

The Devil Wears Prada 2 is a little long and Meryl Streep’s performance lacks the bite that made the first film so memorable. But getting to see Hathaway, Streep, Blunt and Tucci work together again is joyful and escapist.

This film won’t change your life. But it is not trying to. It tells you exactly what it is in the marketing: a celebratory reunion of the actors and a fun retreading of familiar ground. Go for the characters, stay for the nostalgia.

Laura O'Flanagan 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.