The coded gaze

The coded gaze is Joy Buolamwini's name for the patterns of encoded discrimination and exclusion that result when narrow groups of designers and datasets shape the AI systems most of the public interacts with. The framing carries the title of her 2022 MIT Media Lab dissertation, threads the public-facing arc of the Algorithmic Justice League, and is one of the two framings — alongside the evocative audit — that her 2023 book Unmasking AI is built on. Within the corpus, it is the canonical short label for the algorithmic-bias and algorithmic-accountability frame that AJL and its peer organisations work in.

Origin

Buolamwini coined the phrase at the MIT Media Lab in 2016 in the wake of her Aspire Mirror art project, in which off-the-shelf face-detection software failed to register her face until she put on a white mask. The moment crystallised an argument she had been developing about who is and is not encoded into commercial AI systems. The term made its first public appearance in her TEDxBeaconStreet talk "How I'm Fighting Bias in Algorithms," filmed in November 2016, and was set out in print a month later in her founding essay for the Algorithmic Justice League on the MIT Media Lab's Medium (15 December 2016), where she wrote of "individual encounters with bias embedded into coded systems — a phenomenon I call the 'coded gaze.'"

The construction is deliberate. As later commentary has observed, the phrase echoes Laura Mulvey's "male gaze" from 1970s feminist film theory and W. E. B. Du Bois's account of the racial gaze in The Souls of Black Folk; Buolamwini's framing converts those earlier accounts of who looks at whom into an account of who codes whom. The Ford Foundation's 2018 profile quotes her own working gloss: the coded gaze is "the priorities, preferences, and prejudices of those who have the power to shape technology — which is quite a narrow group of people."

Propagation

The framing spread along three tracks that have remained legible since.

Inside AJL, the coded gaze became the umbrella term for the organisation's method work. Buolamwini's 2022 MIT Media Lab PhD dissertation, Facing the Coded Gaze with Evocative Audits and Algorithmic Audits, pairs it with the evocative audit and presents the two as the methodological backbone of AJL's programme. The 2023 book Unmasking AI, in turn, defines the coded gaze as "the evidence of encoded discrimination and exclusion in tech products" and uses it as the through-line for the book's account of AJL's founding, the Gender Shades audits, and the policy advocacy that followed. AJL's own multimedia library catalogues Buolamwini's BBC, New York Times, and Atlantic appearances under the coded-gaze frame.

In philanthropic and civil-society communications, the term was picked up early. The Ford Foundation published "Fighting the 'coded gaze'" in June 2018 as a programme-side framing for its support of AJL-aligned work — an unusually early instance of a major U.S. funder taking a movement framing into its own communications voice. Community and feminist-studies writers reached for it on similar timescales: Pazia Bermudez-Silverman's May 2018 piece in Africana Feminisms on Medium, "The Coded Gaze: Algorithmic Bias?", is one of many class-project and community-syllabus uses of the term that turn up across 2018 and 2019.

In mainstream press and documentary, the framing has tracked closely with Buolamwini's own public appearances. The 2020 feature documentary Coded Bias — in which Buolamwini was the main subject — folds the term into its title and central argument; the November 2023 NPR feature on Unmasking AI uses the coded gaze as the framing for the book's argument that bias is baked into the technology rather than added on top of it.

Why it has carried

Three features explain its travel. First, the phrase compresses an argument that has otherwise been told in long form — about training-data composition, designer-team demographics, deployment contexts, and accountability gaps — into a single noun phrase that fits a headline, a placard, a chapter title, and an academic abstract. Second, it carries an audience-in-the-middle voice: Buolamwini coined and uses the term to address the people on the receiving end of AI systems, not the people building them, and the framing has been picked up most often by writers and organisers doing the same. Third, the term is tied to a method. The pairing of the coded gaze with the evocative audit gives users a noun that names the harm and a verb that names the practice for surfacing it; both halves of that pair have remained legible across AJL's programme work, journalism that draws on it, and the academic-and-policy commentary that has accumulated since 2018.