Il prossimoProssima guida
Generatori di immagini AI in classe d'arte
IA visiva
GUIDA della Società
AI image generators often reproduce and exaggerate stereotypes, for example showing mostly men for 'CEO' or 'engineer', mostly lighter-skinned people for high-status jobs, and narrow beauty standards, because they learn from large, skewed collections of captioned internet images.
Ascolta questa guidaNarrato dall'IA · ~3 min
Researchers measure this by generating many images for neutral prompts and comparing who appears. It matters because these images now fill ads, textbooks and news, and both the bias and clumsy fixes can mislead people.
Text-to-image models such as Stable Diffusion, Midjourney and DALL-E learn the link between words and pictures from huge datasets of image-caption pairs, largely scraped from the web; Stable Diffusion, for example, was trained on subsets of LAION-5B. The web over-represents some countries, languages, body types and occupations, and captions reflect the assumptions of whoever wrote them. The model then learns the most typical picture for a word and tends to amplify it, because generation favors high-probability, 'average' outputs. A prompt that says nothing about gender or ethnicity still produces a default, and that default is often the stereotype. Researchers measure this in a few ways. A common method is to generate hundreds of images for neutral prompts (occupations, adjectives like 'attractive' or 'poor', nationalities) and label perceived gender, skin tone or age, sometimes compared against labor statistics. Hugging Face researchers' 'Stable Bias' project (2023) used clustering of model outputs rather than assigning labels directly, partly because labeling someone's race from a picture is itself fraught. A 2023 Bloomberg analysis of thousands of Stable Diffusion images found that high-paying jobs skewed toward lighter-skinned men and low-paying jobs toward darker skin tones. Fixes are harder than they look. In 2022 OpenAI said it applied a technique so DALL-E 2 produced more diverse people when prompts did not specify, widely reported to involve adding terms to prompts behind the scenes. In February 2024 Google paused Gemini's generation of images of people after it produced historically inaccurate images, such as racially diverse depictions of groups that were historically homogeneous. The lesson is that bias is not a single dial: a system must be representative where the prompt is open and accurate where the prompt is specific. A common misconception is that bias only comes from the training data; model design, filtering, captioning and post-training choices all shape the result too.
A teacher asks an image generator for 'a scientist in a lab' twenty times and gets almost entirely white men in lab coats, then rewrites the prompt to request specific, varied people.
A marketing team generating 'a nurse' and 'a doctor' notices the tool keeps making nurses women and doctors men, and adds a review step before images go into a campaign.
A journalist compares outputs for 'a house in Nigeria' and 'a house in the United States' and finds the model defaults to poverty imagery for one and suburban homes for the other.
A user asks for a historical scene of a specific group and receives an image that inserts demographics that were not historically present, showing how a diversity fix can overcorrect.
Expect more standardized bias benchmarks, more disclosure in model cards, and more context-aware mitigation that distinguishes open-ended prompts from historically or personally specific ones. There is genuine disagreement about the right target: matching current real-world statistics can entrench existing inequality, while uniform representation can misrepresent reality. That choice is a values decision, not purely a technical one, and organizations are likely to be asked to explain it. Users will still need to review generated images before publishing them, particularly in education and news.
Newsletter gratuita
Tre storie verificate sull'IA ogni mattina nei giorni feriali, scritte in un inglese semplice. Gratuito per sempre, senza pubblicità.
Un'email ogni giorno feriale. Disiscriviti con un clic. Non vendiamo né condividiamo mai il tuo indirizzo.
Mettiti alla prova
Feedback immediato su ogni risposta e un certificato condivisibile con un ID verificabile una volta superato un corso.
Sostieni l'istruzione gratuita sull'IA. AI Understanding è un'organizzazione non profit 501(c)(3) — nessun annuncio, nessun paywall, mai. Fai una donazione
AI image generators often reproduce and exaggerate stereotypes, for example showing mostly men for 'CEO' or 'engineer', mostly lighter-skinned people for high-status jobs, and narrow beauty standards, because they learn from large, skewed collections of captioned internet images. Researchers measure this by generating many images for neutral prompts and comparing who appears. It matters because these images now fill ads, textbooks and news, and both the bias and clumsy fixes can mislead people.
Senza alcuna specificazione, il modello produce un valore predefinito ad alta probabilità e tale valore predefinito rispecchia e spesso amplifica gli stereotipi nei suoi dati di addestramento.
Set di dati come LAION-5B contengono miliardi di coppie di immagini-testo raschiate dal web, che portano con sé gli squilibri del web.
I ricercatori generano centinaia di immagini per prompt neutro e analizzano gli attributi percepiti, a volte confrontandoli con le statistiche del mondo reale.
Assegnare la razza in base all’apparenza è inaffidabile ed eticamente problematico, quindi raggruppare i risultati era un modo per confrontare modelli senza imporre etichette.
L’analisi ha rilevato che le occupazioni ben retribuite erano sbilanciate verso gli uomini dalla pelle più chiara e quelle meno retribuite verso le carnagioni più scure.
Continua a imparare
Altre guide selezionate per questo argomento
Il prossimoProssima guida
Generatori di immagini AI in classe d'arte
IA visiva