As agentic AI moves from experimentation into everyday business, education, and public life, one concern is becoming harder to ignore: who is building these systems—and who is being left out?
Science communicator and engineer Dr. Michelle Dickinson has consistently argued that technology reflects the people, assumptions, and cultures behind it. In the era of agentic AI—systems capable of acting autonomously toward goals with limited human oversight—that question becomes even more urgent. Dickinson’s broader work on reducing inequality in technology and making emerging technologies more accessible places her squarely in the conversation about diversity in AI development. (drmichelledickinson.com)
The Rise of Agentic AI
Agentic AI refers to systems that do more than respond to prompts. These systems can:
- Plan tasks
- Make decisions
- Coordinate with other agents
- Adapt to changing environments
- Operate semi-independently over time
Researchers describe these systems as increasingly capable of long-term planning and direct real-world impact. (arXiv)
Businesses see enormous opportunity:
- Automated workflows
- AI-driven research assistants
- Autonomous customer service
- Software-writing agents
- Cybersecurity defense systems
But as the technology grows more powerful, the risks of exclusion and embedded bias also scale.
Diversity Is Not a “Soft Issue”
One of Dickinson’s recurring themes is that diversity in technology is not about optics or public relations—it is about outcomes.
AI systems learn from human-generated data. If the people designing these systems come from narrow social, economic, or cultural backgrounds, blind spots become encoded into the technology itself.
Research continues to show that lack of representation affects performance and fairness:
- AI systems can perform worse on underrepresented groups
- Datasets often reflect historical inequalities
- Design assumptions may exclude certain communities entirely (Artificial intelligence)
In practical terms, this can mean:
- Facial recognition systems failing across skin tones
- Healthcare AI underperforming for marginalized populations
- Recruitment algorithms replicating workplace bias
- Agentic systems optimizing efficiency while ignoring social consequences
Dickinson’s public advocacy aligns with a broader push to ensure that emerging technologies are shaped by a wider range of voices before they become deeply embedded in society.
The Agentic AI Problem
Traditional AI already raised concerns around fairness and accountability. Agentic AI increases the stakes because these systems can act with greater autonomy.
A biased recommendation engine is one thing.
An autonomous system making layered decisions over time is another.
Researchers warn that increasingly agentic systems may create long-range harms affecting marginalized groups disproportionately. (arXiv)
This becomes especially concerning when:
- Training data lacks diversity
- Teams building systems lack representation
- Oversight mechanisms are weak
- Optimization metrics prioritize scale over equity
Dickinson’s approach to science communication often emphasizes making technology understandable and human-centered. That perspective matters in a field increasingly dominated by technical acceleration and commercial pressure.
The Pipeline Problem
The diversity gap in AI remains significant.
Recent reporting on women in agentic AI found that women hold roughly 22% of AI agent engineering roles, with representation dropping sharply at senior levels. (AgenticCareers.co)
This imbalance shapes:
- What products get built
- Which problems are prioritized
- How systems interpret human behavior
- What risks are overlooked
Dickinson has long advocated for widening participation in STEM education, especially among groups traditionally underrepresented in engineering and technology. Her work through science outreach and media has focused on reducing intimidation around technical subjects and encouraging broader engagement with innovation. (drmichelledickinson.com)
The concern is not simply fairness in hiring. It is systemic resilience. Homogeneous teams tend to miss different categories of failure because they often share similar assumptions.
Human-Centered AI
A growing number of organizations are now arguing that successful agentic AI systems must remain human-centered rather than purely efficiency-driven. (McKinsey & Company)
That means:
- Transparent systems
- Explainable decisions
- Inclusive testing
- Diverse development teams
- Human oversight in high-impact domains
Dickinson’s public messaging fits naturally into this direction. Her style avoids both blind optimism and fear-driven narratives. Instead, she frames technology as something people should understand, question, and shape collectively.
That mindset may become essential as agentic systems begin influencing education, healthcare, finance, employment, and governance.
The Risk of Building the Future Too Narrowly
One of the biggest dangers in AI development is speed without reflection.
The agentic AI race is accelerating:
- Governments want strategic advantage
- Companies want productivity gains
- Investors want rapid deployment
But rapid innovation can outpace ethical consideration.
If the people designing autonomous systems represent only a narrow slice of society, then the systems themselves may inherit that narrowness. Diversity becomes a safeguard—not a slogan.
Research increasingly supports this idea, showing that diverse teams improve bias identification, broaden problem-solving, and create more inclusive decision-making processes in AI development. (arXiv)
Conclusion
Michelle Dickinson’s perspective on technology has always emphasized accessibility, inclusion, and practical human impact. In the age of agentic AI, those themes are becoming central rather than secondary.
The future of AI will not be shaped only by technical breakthroughs. It will also be shaped by who participates in building them, who is represented in the data, and whose experiences are considered important enough to include.
Agentic AI may automate tasks, optimize systems, and accelerate innovation. But without diversity in the people guiding its development, it risks reproducing old inequalities at machine scale.
And once autonomous systems become deeply embedded in society, correcting those biases becomes far harder than preventing them in the first place.