Stella最近尝试了一下去英文播客Enrollify串台，聊了聊自己最近两年做的工作，以及对于AI在教育以及高等教育行业应用的看法。也在我们自己的小空间里分享给大家。如果有做类似方向的朋友欢迎来交流！
对Data Science和AI结合领域有兴趣的朋友也欢迎来订阅电子报Data Science x AI:
https://datasciencexai.substack.com/
About the Episode:
In this bonus episode of Higher Ed Pulse, recorded live at the Engage Summit in Charlotte, host Mallory Willsea sits down with Stella Liu, Lead Data Scientist at Arizona State University. The conversation dives into Stella’s journey from tech industry roles at Carvana and Shopify to pioneering ethical AI initiatives in higher education. Packed with insights on building responsible AI tools, this episode is a must-listen for enrollment marketers and higher ed leaders navigating the evolving digital landscape.
Key Takeaways
Ethical AI isn't a luxury—it's essential, especially in high-stakes fields like education.
Accuracy is only one part of AI evaluation; fairness, safety, and bias must be prioritized too.
Small teams without a development staff can still effectively assess AI tools using manual testing frameworks.
Arizona State University's three-part ethical AI evaluation system—Ethical AI Engine, Guard, and Safer—sets a strong precedent for responsible AI integration.
AI should be a solution to a real problem, not a buzzword thrown into tech stacks without purpose.
Effective AI evaluation includes not just internal testing but stakeholder alignment and real-world A/B testing.
What makes ethical AI critical in higher education?
In the conversation, Stella Liu emphasizes that higher ed institutions hold significant responsibility when implementing AI. Unlike e-commerce, where a misstep might mean a delayed package, flawed AI in higher education can deeply impact students' academic journeys and long-term outcomes. Stella argues that with such high stakes, ethical considerations must go beyond accuracy. Institutions must evaluate bias, fairness, and the societal impact of AI-driven tools.She explains that accuracy is just the beginning. AI in education must also pass tests for safety, bias, and transparency. Given that these tools interact with real students, in real time, the margin for error is small—and the consequences of failure, large. Ethical AI, according to Stella, isn't just best practice—it's a moral obligation.
How is Arizona State University tackling AI ethics?
Arizona State has developed a robust three-layer AI monitoring system:
Ethical AI Engine – an automated tool for testing bias, accuracy, and fairness.
Guard – a real-time alerting system to flag anomalies or potential ethical breaches.
Safer – a data analytics tool that audits all AI-user interactions to ensure compliance and integrity.
This framework helps Stella’s team ensure that any AI tool being considered or deployed undergoes rigorous evaluation—both automated and human-led. It's a model other institutions can learn from, especially as AI adoption in higher education accelerates.
Can smaller teams implement AI ethics evaluations?
Yes—and Stella is clear on this point. Not every team has ASU’s resources, but that doesn’t mean ethical evaluations are out of reach. Before developing their current system, her team conducted manual evaluations. They worked with stakeholders to understand real use cases, created test datasets reflective of student queries, and manually scored the AI responses. These efforts, while time-consuming, proved incredibly effective.Stella also recommends A/B testing as a practical tool. By giving one group access to the AI product and comparing their outcomes with a control group, institutions can assess impact and performance with clarity. This approach doesn't require complex infrastructure—just thoughtful design and a commitment to data-driven decisions.
Stella Liu's journey from e-commerce tech to higher ed AI advocacy is a masterclass in translating industry insights into academic innovation. Her experience at Carvana—where she helped develop delivery time algorithms to boost customer satisfaction—mirrors her current mission: solve real problems first, then find the right tech to amplify impact.AI in higher education isn’t just about automation; it’s about building trust, ensuring equity, and enabling smarter, fairer decisions. As Stella perfectly sums it up: "Evolve responsibly—or die."