Achieving consensus: What the path to a code of ethics says about communication
As the usage of AI becomes more widespread, so do discussions about its ethics in the workplace and beyond. Many modern industries, such as the medical industry with the American Medical Association’s Code of Medical Ethics, have established rules for appropriate practice that serve wider society, its clientele, and peers. As the AI industry gains its footing, how might its community come together to arrive at a similar, standardized code of ethics? What takes a discussion in a conference room and turns it into a rulebook that everyone can agree on? Assistant Professor Jessica J. Santana (UCSB), along with co-author Seonghoon Kim (National University of Singapore), tackles this question in their research. Their study reveals that dissenting language homogenizes as discussion progresses, a shift shaped by two primary forces: leadership and the sequence of draft codes.
To come to this conclusion, the authors used a method called semantic network analysis to study email exchanges from the committee behind the 1997 software engineering Code of Ethics. Santana and Kim looked specifically for clusters of words that appeared together in these exchanges. When they viewed these clusters as representations of ethical values, all of which comprised a wider semantic network, the results could be used to measure the number and diversity of ethical values in the represented occupation—in this case, software engineering. What they found mostly matched what has been stated in similar research: that discourse tends to homogenize as practices gain legitimacy. Here, the codification discussion for software engineering progressed from emphasizing licensure, to standardization, and then eventually to education.
What is most striking about Santana and Kim’s findings is their identification of the two driving mechanisms for this shift. First, they noted that the sequential publication of draft codes catalyzed debate, while also condensing semantic communities into larger, but fewer groups of ethical values. Second, they saw that centralized leadership prioritized certain values. In this case, leadership highlighted education as central to the occupation’s ethical identity, after which it ended up as the most prevalent value in the code. This shows that sequential iterations of the code were what opened the door to debate, and led to a decrease in value diversity—and that centralized leadership was what steered the conversation toward those final values.
Why is this so important? On a broader level, software engineering isn’t the only area where policy is created from divergent values. And a code of ethics isn’t the only foundational policy. Research that looks into areas that have been previously overlooked, such as the discussion that predates a code of ethics, can provide a helpful framework with which organizations can work to recognize internal biases and anticipate future occupational challenges.
So, what can organizations do to put these findings into practice?
Santana and Kim’s research emphasizes the importance of both leadership direction and iteration. For leadership, this study shows it plays an incredibly powerful role in shaping the values that its company holds—and beyond it. With this knowledge, preparations can be taken to ensure that branches of leadership are opening the conversation up as much as possible and providing a comfortable environment for feedback. Likewise, with the sequence of draft codes, this research shows that iteration is an effective path to debate. Adaptability to change and a shift toward shared values occurs after multiple meetings and discussions.
With this in mind, here are some action items that organizations may consider implementing, both during institutionalization and after it:
- Involve leadership with other team members as much as possible. With this research showing how much sway leadership’s values can have on decision-making, it is essential that leadership understands their team members’ priorities and is willing to include them in the conversation.
- Prioritize management trainings for leadership. Once again, with the amount of influence leadership is able to have on foundational policy and other projects, leadership needs to be trusted and have the skills needed to spearhead discussion. These trainings should both provide leadership with these tools, and a thorough understanding of this responsibility to their organization.
- Set multiple meetings and reviews for important projects. As shown in this study, leadership leads to more conversation, debate, and understanding, ultimately pushing team members toward more common ground. Having an entire team on the same page for any project is a powerful way to increase efficiency.
- Provide an environment for feedback on organizational policy. A text like a code of ethics may seem like an end-all-be-all, but this study shows that it still results from human conversation. A code can only improve and lean towards shared value with conversation and feedback.
Instagram
LinkedIn
Twitter