Today, one ethics course simply isn’t sufficient for the engineer working with AI. It’s the blunt inlay of a broad trend in engineering education that is now starting to sweep across the discipline, in which the traditional combination of equations, coding, and lab technique seems to be coming to an end and systems are recommending, ranking, diagnosing and deciding at scale. Many programs continue to educate students for a world that’s gone.Many programs are still educating students for a world that has been profoundly changed, says Karen Panetta, IEEE Fellow and dean of graduate education at Tufts University’s School of Engineering. Her perspective is that the discussion of whether or not engineers will adopt AI is quickly evolving to whether they will grasp its constraints sufficiently to safeguard humans from its failures.
It is relevant to Panetta’s critique to begin with. “Nobody’s interested in looking at the nitty-gritty tools,” she said. Now it’s all about the impact connection to the STEM fields. That’s a pain at the butt, but it’s also a technical change. Increasingly, engineers create systems for communities, such as the homes they live in, the workplaces they go to, the hospitals they attend and the schools they go to, in which errors in design have directly tangible consequences. An idealized representation in the laboratory can yield skewed and/or unsafe results in the real world if the results are based on a population that does not include the individuals being represented.
It’s captured in education technology alone. OECD reviewed research studies found evidence of algorithmic bias in assessment tools for spoken language, essay grading, predicting student dropouts, and identifying when students are at risk of dropping out of their courses. These mistakes aren’t caused by a single line of code. They can manifest in the data collection, data labelling, data evaluation or deployment and they are usually not seen until a system is impacting real students. It’s the reason that ethical training is increasingly being taught to work with fair/ethical aspects of design, not as de-Brief sessions at the end of the build.
Again at a simpler level, Panetta states that. It cannot be taught in just one course and then carried forward to every other course, she said. RIGHT ACTION has to be done consistently and synergistically across all of the curriculum. This seems to fit into a broader standards push for trustworthy AI. IEEE has added to its set of work on transparency of autonomous systems, algorithmic bias, privacy processes, explainability, dataset quality and organizational AI governance. On the diffusion of those standards, it is indicated that supervision of ethics is heading towards closer proximity to the engineering practice itself. Turning it into proceeding, provable, otherwise depending into system design decisions, and not only skilled ideals.
No, Panetta is not saying that we shouldn’t use AI. Of her students, she said that she wanted them to use AI. But we’re afraid they’re going to get carried away with it, said he. Her worry is about software dependence, less about voice and judgment. Not that it would be their voice, we don’t want that be the case, she said. It is my desire to discover and hear their voice. In engineering, that means AI has been put to good use at aiding the drafting and the analysis and exploration of these systems, but doesn’t supplant accountability when it comes to careers, to health, to safety.
The other one is to squeeze the gap between university and practice. She states that short internships don’t give students a whiff of the real world of engineering, where compromises, miscommunication and user action can influence technical choices. She prefers longer industrial experience, mentorship and instruction that encourages free-form problems solving over presentation. She said, Don’t tell me what it is in the book! Share with me how you would do it! Perhaps this is the actual learning change for kids. As algorithms come to shape society, engineering schools are being compelled to teach both the inner workings, but also how to ask questions about that, before others face the consequences.