INTRODUCTION
Artificial intelligence (AI) has rapidly infiltrated nearly all aspects of society. From driving our cars to managing our schedules, from generating code to serving as emotional companions, AI’s influence on our lives is steadily expanding.
However, one domain remains largely untouched by the reach of AI: moral reasoning in leadership. As AI continues to reconfigure our lives, its potential to influence, guide, or even reshape our moral reasoning and organizational leadership practices warrants scrutiny. We live in an era where leaders regularly grapple with complex and morally charged challenges. Traditional approaches to addressing these challenges lean on human judgment and mentoring, with leaders often seeking counsel from trusted advisors or mentors to tackle these dilemmas. Could AI fill these shoes?
The HOW Institute for Society, committed to fostering a culture of moral leadership and principled decision-making in an ever-evolving world, sought to investigate this very possibility. We carried out research involving interactions between leaders and ChatGPT, the much-discussed language model developed by OpenAI. This research brief presents our findings, exploring the benefits and limitations of AI as a source of guidance through moral leadership challenges.
Our goal is to spotlight this critical yet overlooked aspect of AI’s potential, shedding new light on its role in moral leadership. In doing so, we hope to spark an important conversation about this crucial intersection of technology and ethics, in alignment with the Institute’s mission to elevate humanity in the face of the profound social, economic, and technological disruptions of the 21st Century.
RESEARCH
APPROACH
The HOW Institute for Society’s flagship program is the NXT-GEN Fellowship for Moral Leadership, which serves as an incubator for emerging managers and leaders tasked with navigating our interconnected and morally charged world. The fellowship offers participants the tools, language, and mentorship necessary to inspire and amplify values-based behavior in an increasingly complex workplace.
Seeing the unique potential of this group, we invited NXT-GEN alumni to share leadership challenges they had recently faced at work, providing a database of real-life scenarios for our investigation. The narratives shared by our NXT-GEN alumni provided a vivid tableau of leadership challenges spanning an array of situations and industries – from navigating office politics and managing inter-departmental tensions to addressing concerns of racial and gender bias. One leader grappled with the fallout of a poorly communicated internal reorganization, another navigated the ethical lines blurred by ambiguous expense tracking directives. The struggles of merging multiple legacy businesses, balancing organizational needs against employee welfare during times of austerity, and addressing a worrying trend of employee disengagement cropped up in multiple accounts. The breadth of these challenges served as a testament to the multifaceted complexities that today’s leaders face. They also offered fertile ground for exploring how an AI like ChatGPT might offer insight and guidance in navigating these ethically and emotionally fraught terrains.
The core of our research comprised a series of in-depth conversations with 12 NXT-GEN alumni in which we presented ChatGPT with the leadership challenge the participant previously shared with the HOW research team. However, before conducting these interviews, we realized the importance of familiarizing ourselves with ChatGPT’s operational characteristics to ensure that we could effectively mediate conversations between the research participants and the AI. To gain a thorough understanding of how ChatGPT behaves, we conducted a beta testing phase where we presented the AI with leadership challenges experienced by The HOW Institute staff at their past employers.
The results of our beta testing phase left us underwhelmed. Although the AI demonstrated an adequate understanding of the challenges we presented during this initial phase of our research, it jumped into giving recommendations without fully exploring the context of each challenge. The AI’s behavior was reminiscent of a search engine, promptly delivering information but lacking the depth and nuanced understanding typically expected of a competent and caring advisor.
Recognizing that the AI needed guidance to transcend this search engine-like behavior, we endeavored to train it to act more like a coach. We focused on fine-tuning our prompts, encouraging ChatGPT to engage in a conversational manner, asking questions to gain a fuller understanding of the challenge and the stakes involved before launching into recommendations. Through trial and error, we converged on a prompt that better captured the essence of the role we wanted ChatGPT to play:
Armed with a deeper intuition for ChatGPT’s behavior from this beta phase, we were set to embark on interviews with the NXT-GEN alumni. Mediating an exchange between each alum and the AI, we invited the former to seek advice from ChatGPT as if they were encountering the problem for the first time. We followed up each interaction with a debriefing session in which we inquired about their perception of ChatGPT’s understanding, the helpfulness of its input, their surprises and disappointments during the interaction, and their likelihood of using such a tool when facing future leadership challenges. Further, we asked about their views on the advantages and pitfalls of relying on these AI tools.
NAVIGATING
LEADERSHIP CHALLENGES
Oriented by the refined prompt, ChatGPT was surprisingly adept at processing and advising on the leadership challenges that we gave it. In general, after one or two interactions, the interviewees felt that the bot understood the scenarios they presented. One interviewee stated that explaining their challenge to the bot was “quicker than the several conversations I needed to have with HR to explain the situation.” Another commented that they were “shocked” by how well ChatGPT performed.
Several features of ChatGPT’s behavior are worth pointing out here. First, it communicated with a sense of empathy that, even if artificial, made the interactions feel more authentic. It also did a good job of collecting information before jumping to recommendations (a problem we had observed before arriving at our finalized prompt). For example, it responded in the following way to an interviewee’s description of suddenly being treated harshly by their manager:
Our interviewees appreciated this seemingly empathetic and curious orientation, with one stating that “I realize there can’t actually be an emotional connection because it’s ChatGPT, but I feel that the questions that it asked to get more clarity were coming from a place of wanting to help. I felt listened to.”
We also found that ChatGPT’s advice was often thorough and actionable. Several interviewees went into the experiment expecting the bot to offer only management platitudes but were pleasantly surprised by the practical specifics it gave. “It was quite detailed,” noted one interviewee, who requested the transcript be sent to her so that she could use some of the direct language produced through the interaction. “It was giving me a true framework with actual clear steps that I can take, not just a high-level statement.”
AI tools can analyze interpersonal skills and provide data-driven insights for talent development, thereby enhancing leadership effectiveness.
For example, presenting a challenge involving a lack of collaboration between two internal teams, one participant asked the bot how to organize a cross-team meeting without coming across as confrontational.
The bot responded with the following:
When given appropriate guidance, ChatGPT provided relevant and insightful advice on a variety of leadership challenges. It demonstrated an ability to grasp the situation, ask relevant questions, and offer detailed, actionable recommendations. Our NXT-GEN alumni found value in ChatGPT’s counsel, recognizing its potential as a tool for leaders grappling with intricate moral and ethical dilemmas. Indeed, almost all of them expressed a high likelihood of using ChatGPT or a similar type of tool when they encounter new leadership challenges in the future.
But they also pointed out several inherent problems in relying exclusively on ChatGPT for direction.
AI AS A DECISION-MAKING AID
Despite being clearly impressed by ChatGPT’s abilities to process and advise on their leadership challenges, our interviewees viewed the bot as a supplement to—not substitute for—their own leadership intuition. As one interviewee put it, “the bot presented a list of options, but in the end, I still need to make the call.” Another interviewee explained that “ChatGPT doesn’t understand all the nuances of your scenario and your organization, and so that’s where your instinct comes in. You need to decide what would work or not, based on what you’ve seen and on your own knowledge.”
The primacy of the interviewee’s leadership intuition became particularly apparent when they pushed back on what ChatGPT suggested. For example, one participant presented a challenge in which they were struggling to delegate responsibility to a direct report because of a conflict between their values of fostering autonomy and of maintaining high-quality deliverables. ChatGPT’s initial response included a recommendation to convey to the direct report the need for maintaining the team’s reputation. Our interviewee responded by noting that making the direct report feel responsible for the team’s reputation could create undue pressure, leading ChatGPT to respond with the following:
It is important to recognize that ChatGPT effectively pivoted to a more promising strategy only because of deliberate and thoughtful guidance from the human interlocutor. Reflecting on this experience, this interviewee explained that “asking these questions of ChatGPT without having done the work to understand your own moral values would be problematic.” In other words, having a seemingly knowledgeable and thoughtful AI chatbot at one’s disposal is no replacement for the reflection and study required to develop one’s own moral leadership intuitions.
Indeed, our research participants consistently found ChatGPT’s advice to be a source of useful input for decision-making, but not something to rely on exclusively. “It can be another voice,” said one interviewee, “but you have to gut check it to make sure the response is appropriate for the situation.” Another interviewee pointed out that “ChatGPT doesn’t understand the dynamics or the culture of your scenario and your organization. And so that’s where your instinct comes in to say ‘is this something that would work here or not, based on my own knowledge?’” Several interviewees insisted that getting advice from trusted human colleagues will remain central to their strategy for tackling future leadership challenges. “I still value human interaction,” said one interviewee. “I still would probably ask someone first, and then I would talk to ChatGPT just to see what the responses are – as a gut check.”
While some participants stated that they would consult ChatGPT as a final check, others suggested that consulting AI might be a good first step in the decision-making process. “Having someone or something to talk through the issue with could help me to clarify my own thoughts,” reflected one interviewee. Along similar lines, another interviewee said, “asking the question to ChatGPT could be a good way to implement a pause – before I say or do anything.” As several of our participants acknowledged, using AI as a tool for self-reflection and clarification has the advantage of efficiency (the bot is available at a moment’s notice) and of confidentiality (there is no judgment for half-baked thoughts or sensitive opinions). As one interviewee put it, “it’s a good outlet for getting your feelings out.”
AI AND THE FUTURE OF MORAL LEADERSHIP
Our research shows that the immediacy, efficiency, and confidentiality offered by generative AI tools such as ChatGPT can help us navigate morally charged leadership challenges. However, it is imperative to recognize that these tools lack the innate human qualities that are essential for truly impactful leadership: intuition, lived experience, empathy, and a deep understanding of context. AI models such as ChatGPT operate within the confines of their programming and the data they have been trained on. They lack the intricate tapestry of personal experiences, values, emotions, and exposure to unique organizational dynamics that a human leader possesses.
The true challenge, then, lies in leveraging AI as a co-pilot on this journey without letting it take control. This delicate balance ensures that we harness the best of both worlds—the rapid analytical capabilities of AI and the rich, morally-guided intuitions of human leaders.
The principles we have distilled from our study are crucial guidelines for moral leaders, and to organizations that value moral leadership. We detail these guidelines below.
Detailed and Deliberate Interactions: Always frame your interactions with AI tools like ChatGPT with precision. The clarity and richness of your prompts determine the quality of the feedback. Before querying, outline your dilemma or challenge in its full context. The more background and detail you provide, the more nuanced and tailored the AI’s response will be, optimizing the relevance of its advice.
Retain Human Primacy in Decision Making: Use AI as a guide, not a guru. No matter how seemingly insightful the AI’s suggestions, always overlay them with your understanding, intuition, and hard-earned expertise. You bring a wealth of contextual knowledge and experience to the table. Your perspective and judgment, honed through real-world experiences, should always have the final say.
-
Continuous Moral Development: Dedicate time to tweaking your moral compass. Engage in reflective practices, attend workshops, read widely, and actively discuss ethical dilemmas with peers. An AI tool can provide options, but only a well-developed moral intuition can discern the right path. Organizations should provide avenues for employees at all levels to build their moral leadership skills. As our founder Dov Seidman explains, “Leaders don’t stop learning and growing just because they’ve accumulated formal authority in an organization. They continue to build moral muscle—I call it ‘going to the moral gym’—by wrestling with questions of right and wrong, fairness and justice, what serves others and what doesn’t.”
-
Human Mentorship Remains Vital: Always consider human advisors as a primary source of counsel. AI tools are valuable, but they should complement, not replace, human mentors, who offer a depth of understanding, experience-based wisdom, and emotional support that AI cannot replicate. They can provide tailored guidance, taking into account unique interpersonal dynamics and organizational cultures. If you are filling the role of a mentor or advisor, do not assume that an AI tool will be able to provide your employees guidance with which they feel completely comfortable.
