Garbage In, Garbage Out: The Data Problem Hiding Inside Your AI Strategy - with Caroline Jarrett

In this episode of 'This is HCD', host Gerry Scullion welcomes Caroline Jarrett, a renowned forms and survey design expert, to discuss the critical but often overlooked issue of data quality in AI strategy. Drawing from her three-decade career in human-centered design, Jarrett emphasizes that 'garbage in, garbage out' isn't just a cliché—it's a systemic failure in how organizations treat data, especially in forms and surveys. She argues that forms are the only compulsory part of any service, yet they're frequently designed without regard for usability, accuracy, or error rates. Jarrett reveals that many organizations don't even know their own error rates, and that AI adoption without fixing foundational data problems risks amplifying existing flaws. She introduces a framework of six types of service errors—from wrong results to delayed impacts—and advocates for measuring data quality through visualization, analytics, and observational research. The conversation underscores the urgent need for designers to lead data quality conversations before AI is implemented at scale. Key takeaways include: 1) Forms are the most visible and compulsory part of any service, yet are often poorly designed; 2) Organizations must measure and understand their error rates before adopting AI; 3) Data quality issues like inconsistent entries, missing values, and invalid inputs (e.g., apostrophes in names) are widespread and systemic; 4) Visualizing data helps designers spot anomalies and inconsistencies more effectively than raw numbers; 5) AI should not be used to 'fix' bad data—it will only magnify it. Jarrett concludes by urging designers to use the promise of AI as a lever to push for better data hygiene, turning technical debt into strategic advantage.