ENGINEERING HEARTS® RESEARCH
AI Ethics Experts discussing the challenges of ethics integration in
computer software and engineering education.
The Twitter sphere is not science, obviously, but it presents an interesting opportunity to watch in "real time" the global conversation on the struggle to embed ethics into engineering and engineering education. Observational research here is seeking to explore how ethics is discussed in this context. My aim is to briefly present this one social perspective to illuminate how this population of stakeholders, including myself, engage and exchange ideas on AI ethics as a subset of ethics in engineering.
After reading Author David Gunkel and Professor Shannon Vallor's exchange on the failing of CS/ENG education to integrate ethics into higher edu CS/ENG curricula; I purposefully moved the conversation from education to engineering practices. If the suggestion is that ethics is not being "taught" in an integrated manner across engineering curricula, what did these folks think about embedding it into requirements engineering. Hence my research question: "Where/when would you embed "ethical value requirements" in the software and systems engineering processes?
I've been in the IEEE P7000 Working Group for over 3 years. We are developing the "Model Process for Addressing Ethical Concerns During Systems Design". This is our working definition of "ethical value requirement (EVR): organizational or technical requirement catering to values that stakeholders and conceptual value analysis identified as relevant for the SoI (system of interest).
Professor Alan Winfield and others were kind enough to answer the Twitter poll. 1001 votes were received during a 7 day polling between July 22nd and July 29th, 2020. Below are the surprising results. Based on our 3 year long conversations and standards development work at IEEESA among a tight knit group of about 50 software and systems engineers and other domain experts; I expected the results to favor the first option:"During design." I was pleasantly surprised most of these Twitter participants (32.8%) voted to embed EVR's throughout the lifecycle with continued iterations. Quite the opposite of my expectations in fact. The "During design" option received the least proportion of votes. So this poll would suggest that most of the Twitter participants think EVR's should be embedded during design, development and indeed be embedded throughout the software and systems engineering process with iterative lifecycle integration. Fascinating! I will follow up on this social media research with a formal survey using more scientific methods. Stay tuned.
Where would you embed ethical value requirements?
The above Twitter research was a small window into how AI Ethics experts perceive embedding ethics into requirements engineering. Below, I selected two primary engineering text books and picked out their requirements engineering approaches to highlight their differences and similarities. For now, these are placeholders in my thinking on how engineers are formally taught to consider the design space, the challenges and opportunities.
Two perspectives below on requirements engineering.
#1
Requirements Engineering for Software and Systems (Applied Software Engineering Series) 3rd Edition (2017) LaPlante

#2
Mastering the Requirements Process: Getting Requirements Right (2012) Robertson & Robertson
