Building human-centered AI products with Ovetta Sampson

Key Concepts:

• Human-centered AI design
• AI ethics and responsibility
• Democratizing AI
• ML/AI infrastructure
• Cognitive biases in AI
• Traumatized data sets
• Augmenting human capabilities with AI
• AI literacy
• Limitations of AI

1. Ovetta Sampson's Origin Story and Early Career:

• Ovetta's initial exposure to technology began at age nine when she started learning programming on a Commodore 64, inspired by her father's COBOL and Fortran studies.
• Simultaneously, she developed a passion for writing, having her first story published and actively participating in writing and storytelling activities in school.
• In high school, she worked as a journalist and designed the school newspaper, gaining experience with design programs like Quark and Unisys. This involved a blend of creativity and programming to fit content onto pages.
• As a freshman in college, she joined "The Index" newspaper and later became the sports editor, designing large special sections and working in radio and television.
• She describes her brain as "hardwired to translate creativity into tools" and vice versa, emphasizing the interplay between creative expression and technological implementation.

2. Transition to Web Development and Data-Driven Journalism:

• After college, Ovetta joined "St. Joseph News-Press" in 1994, where she began transferring newspaper content to the web, marking her entry into web development.
• She later worked as a crime and education reporter, utilizing computer-assisted reporting techniques that involved number crunching and analyzing large datasets.
• To make sense of the data, she learned programming languages like SQL, R, and Python.
• She took predictive analytics and statistics classes in grad school, realizing she was the only designer in the room. This led to her discovery of machine learning.

3. Human-Centered AI Design at IDEO:

• After graduating with a master's in computer science, Ovetta joined IDEO in 2016 and worked with DataScope to explore human-centered AI design.
• She was the first design researcher embedded with DataScope, working on projects to integrate human-centered design with machine learning and AI systems.
• Example: A project for a medical malpractice company aimed to prevent lawsuits by upskilling doctors. Ovetta emphasized considering the consequences of predictive models on doctors' reputations.
• She mapped the journey of a malpractice suit and identified a woman who worked at the company and was able to predict which cases would result in malpractice. This woman was called "the human algorithm" and her insights were used to build the model.
• Ovetta emphasizes that human-centered AI design is about augmenting human capabilities rather than solely focusing on technology.

4. The Importance of Human-Centered AI:

• Ovetta argues that much of AI today is tech-driven rather than human-centered, focusing on the capabilities of the technology rather than solving human problems.
• She contrasts this with human-centered AI, which identifies human needs and uses technology to scale and augment solutions.
• Example: Benjamin Cabé's bread-proofing AI nose, which was then adapted by a 13-year-old to create a pneumonia-detecting nose.
• Ovetta cautions against technology for technology's sake, arguing that it can have negative consequences if not guided by human-centered principles.
• She emphasizes that technology is not agnostic and has sociocultural effects on society.

5. Cognitive Biases and Traumatized Data Sets:

• Ovetta discusses the dangers of using traumatized data sets in AI models, citing the example of Chicago's strategic subjects list, which used police arrest records to predict potential crime victims and perpetrators.
• She highlights the issue of using race as a way to identify suspects, noting its unreliability.
• She researches how AI intersects with cognitive biases to produce negative outcomes.
• Ovetta argues that it's essential to acknowledge the potential harm of technology and design for it.

6. Responsible AI and Ethical Considerations:

• Ovetta challenges the notion that there is a tension between business models, responsible AI, product design, and AI capabilities.
• She argues that these aspects should be considered equally important.
• She advocates for incorporating ethics education into computer programming curricula to encourage critical thinking about biases.
• She emphasizes the importance of acknowledging that AI can go wrong and developing avenues for users to report and address issues.

7. Democratizing AI and ML Infrastructure:

• Ovetta defines democratizing AI as making it accessible, open, and human-centered.
• Her team works on ML infrastructure to build tools and platforms that developers use to create AI applications.
• They aim to shift the burden of responsible AI from individual developers to the system itself, alerting developers to potential issues like biased data sets.
• She draws a parallel to the banking industry, where models are audited to prevent discrimination.

8. The Evolving Definition of a "Person of AI":

• Ovetta discusses the evolving definition of a "person of AI," from researchers and developers to anyone using AI.
• She shares an example of a high school coach who used AI to clone his principal's voice and create a racist recording, highlighting the potential for misuse.
• She emphasizes the importance of kindness, empathy, and using technology for the benefit of others.
• She advocates for AI literacy to help people discern fake from real information.

9. Focusing on Limitations and Designing for Safety:

• Ovetta urges a shift in focus from the capabilities of AI to its limitations.
• She emphasizes the role of human-centered designers in ensuring that AI does not cause cognitive, physical, or psychological harm.
• She concludes by emphasizing the need to shape AI, teach it what's good and bad, and prevent its weaponization.

10. Synthesis/Conclusion:

Ovetta Sampson's insights emphasize the critical need for human-centered design and ethical considerations in the development and deployment of AI. She advocates for a shift in focus from technology-driven AI to AI that augments human capabilities and addresses real-world problems. By incorporating ethics education, addressing cognitive biases, and designing for the limitations of AI, we can create more responsible and beneficial AI systems that serve humanity. The key takeaway is that AI is a tool, and like any tool, its impact depends on the intentions and values of those who wield it.