Ignoring D&I in AI systems can cause digital redlining, discrimination, and algorithmic oppression, leading to AI systems being perceived as untrustworthy and unfair.

The issue manifests across the full AI lifecycle, from who builds systems, to what data trains them, to who benefits from their outputs. It encompasses:

• Biased training data that over-represents dominant demographic groups and under-represents others, causing systems to perform worse for minorities, women, the elderly, or disabled people.

• Homogeneous development teams that lack the diversity of perspective needed to anticipate how systems affect different communities.

• Inaccessible design that excludes users with disabilities, non-native language speakers, or those with limited digital literacy.

• Disparate impact where ostensibly neutral systems produce systematically unequal outcomes across protected groups.

• Language technology bias, where if these technologies are biased against certain genders or reinforce harmful stereotypes, they can perpetuate inequality and exclusion — particularly evident in machine translation and multilingual scenarios.

Why it matters

AI and automation systems are increasingly making or informing decisions in high-stakes domains such as hiring, healthcare, criminal justice, housing, education, and credit, that shape life opportunities for billions of people. 

When these systems are built without diversity and inclusivity at their core, they risk encoding historical inequalities into infrastructure that operates at scale and speed far beyond human review, making discrimination both more systematic and harder to detect or challenge. 

Impacts & consequences

The consequences of ignoring of downplaying diversity and inclusion in AI and automation systems can be significant, including:

• Wrongful exclusion: People from minority and vulnerable groups such as the disabled may be denied jobs, housing, credit, or services by systems they cannot see or contest.

• Reinforcement of stereotypes: Generative AI tools have come under scrutiny for reinforcing both gender and racial stereotypes. Image generation tools repeatedly produce visuals of professions like "judge" or "CEO" showing mostly white males, despite demographic diversity in those roles. 

The neglect of D&I can also lock in historical bias, leading, over time, to deepening inequality, reduced public trust, and the normalisation of unfair treatment by automated systems.

Causes & contributing factors

The neglect of D&I can stem from various sources: 

• Homogeneous development teams. A lack of diverse perspectives and lived experiences in engineering and data science teams leads to unnoticed algorithmic biases.

• Biased training data. A reliance on legacy datasets that contain historical prejudices and underrepresent minority demographics.

• Data exclusion. Filtering out "edge cases" or minority populations to simplify model training, resulting in poor performance for those groups.