In the 2010s, John Gallagher, now an associate professor of English at the University of Illinois Urbana-Champaign, embarked on an unexpected journey into the world of artificial intelligence. During his doctoral studies, he needed a faster way to evaluate thousands of comments for his dissertation. This necessity drove him to explore machine learning tools, igniting his interest in the convergence of AI and communication.

By 2022, Gallagher's initial curiosity evolved into a comprehensive research project, funded by the Notre Dame-IBM Tech Ethics Lab, to investigate the communication challenges faced by AI and machine learning academics. His study was ambitious: over two years, Gallagher conducted over 100 interviews, predominantly with PhD holders across various fields, focusing on those teaching and researching AI and machine learning in higher ed. The result was nearly 100 hours of audio recordings, analyzed by Gallagher and his graduate students to uncover recurring themes and insights.

The study's findings reveal several communication breakdowns:

• Managing hype: The scholars frequently contend with misconceptions driven by hype. Often, news coverage amplifies inflated expectations and unrealistic fears. The findings reveal widespread public misunderstandings about AI's potential, largely influenced by the "Hollywood effect" from movies like "The Terminator."

• Dealing with publication pressures: The academic landscape is fiercely competitive, exacerbated by platforms like arXiv that hasten information dissemination. PhDs are pressured to produce more papers faster, often at the expense of thoroughness and rigor. Major conferences such as NeurIPS, ICML, and ICLR dominate the field, compelling AI and machine learning academics to move from one deadline to the next, resulting in less refined and error-prone publications.

• Communicating technical knowledge effectively: Clear communication of academic expertise is essential yet challenging. PhDs must balance detailed, accurate explanations with accessibility for non-experts. Terminological ambiguities frequently cause miscommunication, particularly in interdisciplinary work where pedagogical terms have different meanings.

• Clarifying misconceptions: AI is a broad field dedicated to creating intelligent machines, while machine learning is a specialized subset focused on developing models that learn from data. A PhD in AI usually emphasizes theoretical and foundational principles, whereas a PhD in machine learning is oriented toward practical applications of AI. Understanding the distinction between the two is essential to avoid misconceptions.

Moving Forward

The research underscores the importance of effective communication in AI and its rapidly evolving subset, machine learning. By addressing the challenges of hype, publication pressures, and technical barriers, the study offers valuable insights into how PhDs can better convey the complexities of their work and its public perception. As AI and machine learning continue to advance, facilitating accurate communications will be essential in shaping a future where its capabilities and limitations are better understood.

Learn More

You can read a more detailed account of Gallagher's research or watch Explicit & Tacit Ethics from Machine Learning Researchers, an online training module that he developed from over 100 interviews to improve AI and machine learning professionals’ writing and documentation skills. 

...

The Notre Dame–IBM Technology Ethics Lab, which is a key element of the Institute for Ethics and the Common Good and the  Notre Dame Ethics Initiative, promotes broad-based, far-reaching interdisciplinary research, thought, and policy leadership in artificial intelligence and other technology ethics by engaging with relevant stakeholders to examine real-world challenges and provide practical models and applied solutions for ethical technology design, development, and deployment. The Lab is sponsored by IBM through a 10-year, $20-million investment.

Originally published by Tech Ethics Lab at techethicslab.nd.edu on July 02, 2024.