DalšíDalší průvodce
COMPAS a zkreslení v algoritmech recidivy
Společnost
PRŮVODCE společnostmi
Twitter’s 2021 audit of its image-preview saliency model found small selection disparities in a controlled paired-image experiment: an 8% difference from parity favoring women, 4% favoring white over Black people, and 7% favoring white over Black women.
Poslechněte si tento návodAI-vyprávěný · ~3 min
The company found no systematic “male gaze” cropping in its separate 100-image-per-group check and replaced automatic cropping with full-image previews.
Twitter used a saliency model beginning in 2018 to choose a focal point for image previews. The model was trained on human eye-tracking data, assigned saliency scores to image regions, and centered a crop on the highest-scoring point. That convenience also took crop control away from authors. In 2020, users posted examples where Black people or women appeared excluded from previews, prompting Twitter to publish an initial analysis and later a larger audit. In May 2021, Twitter described a paired-image experiment. The team randomly linked images of individuals from different groups and measured how often the system selected each image as the salient one, with 50% as its demographic-parity reference. It reported an 8% difference from parity favoring women over men, 4% favoring white over Black people, and 7% favoring white over Black women; its reported Black-versus-white men difference was 2%. These are results under that test design, not a census of all uploaded images, a causal account of training data, or a measure of all possible user experience. The company separately tested the “male gaze” concern using 100 male-presenting and 100 female-presenting images that had multiple salient regions. It found about three per hundred in each group cropped away from the head, usually to nonphysical image details, and reported no significant objectification bias in that sample. The sample was small and the company explicitly identified other possible harms, including cultural nuance and reduced user agency. Twitter launched full-image previews for standard aspect ratios on mobile after testing; the May 2021 post says the feature rolled out to everyone. It characterized image cropping as a decision people should make. Do not say Twitter trained the model to recognize race: it predicted saliency from image patterns, then audit researchers compared selections across demographic groups.
A researcher reproduces the published paired-image test and reports its selected comparison groups and parity baseline.
A product team lets image authors preview and control a crop instead of inferring the focal point for them.
A reviewer explains why the saliency model’s output is not a race-recognition label or a measure of viewer attention in every context.
A journalist separates Twitter’s 2020 audit results from its May 2021 expanded analysis and subsequent display change.
The case illustrates both statistical assessment and product redesign. Twitter removed saliency-based cropping for standard image previews after user criticism and its internal analysis. The company’s 2021 post is historical evidence about that feature and response; it does not establish that every later X image-display surface uses the same code or policy. Evaluate deployed versions and real user controls separately from old benchmark results. New audits should prespecify comparison groups, include image creators’ preferences, and publish uncertainty so that parity metrics do not obscure representation harms.
Newsletter zdarma
Tři ověřené příběhy ze světa AI každý všední den ráno, psané srozumitelnou angličtinou. Navždy zdarma, bez reklam.
Jeden e-mail každý všední den. Odhlásit se můžete jedním kliknutím. Vaši adresu nikdy neprodáváme ani nesdílíme.
Otestujte se
Okamžitá zpětná vazba ke každé odpovědi a sdílitelný certifikát s ověřitelným ID po absolvování kurzu.
Podpořte bezplatnou AI vzdělávání. AI Understanding je nezisková organizace 501(c)(3) — žádné reklamy, žádná platba za přístup, nikdy. Darujte
Twitter’s 2021 audit of its image-preview saliency model found small selection disparities in a controlled paired-image experiment: an 8% difference from parity favoring women, 4% favoring white over Black people, and 7% favoring white over Black women. The company found no systematic “male gaze” cropping in its separate 100-image-per-group check and replaced automatic cropping with full-image previews.
Twitter said the model learned saliency scores from human eye-tracking data.
Twitter described scoring image regions and using the highest-scoring point as the crop center.
Twitter reported a 4% difference from demographic parity favoring white individuals in paired Black/white comparisons.
The company sampled 100 male-presenting and 100 female-presenting images and found about three out-of-head crops per hundred in each group.
Twitter said it launched full-image display for standard aspect-ratio photos after testing.
Učte se dál
Pro toto téma bylo vybráno více průvodců
DalšíDalší průvodce
COMPAS a zkreslení v algoritmech recidivy
Společnost