| Abstract: |
Humans are not born with all the knowledge or skills needed to navigate the
countless decisions, challenges, and uncertainties they face every day in
their personal and professional lives. Instead, we rely on our ability to
learn, developing strategies and acquiring competencies that allow us to adapt
to our environment and grow over time. While some knowledge can be acquired
through personal experimentation, many of the abilities we depend on have
become too complex to be rediscovered through trial and error alone. As a
result, humans have long recognized that progress, whether scientific,
artistic, or otherwise, is only possible because we build on the insights and
experiences of others. Put differently, we often learn far more effectively
when we draw on existing knowledge rather than by “reinventing the wheel.”
Through direct observation, imitation, collaboration, and instruction, people
can acquire knowledge more quickly and with fewer risks than they could
through solitary exploration. The same is true for indirect forms of social
learning, such as reading books or searching the internet, which provide
access to the cumulative experience of others across both space and time. As
knowledge accumulates faster than any person can master it, technologies that
help us acquire, store, and share information have become indispensable to
organizational and societal progress. Artificial intelligence (AI) represents
one of the latest milestones in a long line of technologies that have
transformed the way people learn. However, unlike earlier breakthroughs such
as the printing press or the internet, AI’s impact goes beyond reshaping how
we store and share existing knowledge. With its ability to discover complex
relationships in large amounts of data, AI can provide advice, make
recommendations, and even contribute its own knowledge to organizational and
societal learning processes. For example, AI can help discover promising
cancer treatments by predicting patient responses to different immunotherapy
drugs. This transformative potential has further grown with the emergence of
generative AI, which can create entirely new content such as text, images,
music, and videos by recombining and building on the human knowledge embedded
in its training data. Rather than passively augmenting human learning, AI is
actively collaborating with humans in the process of knowledge creation. As a
result, few doubt that AI will continue to influence what we learn, how we
learn it, and even who we learn from. However, this impact is unlikely to be
uniform across all contexts. While AI has the potential to accelerate
discovery and democratize information access, it may also obscure human
expertise or reinforce existing biases. To guide researchers and practitioners
in integrating AI in ways that augment, rather than displace, human learning,
this dissertation examines the impact of AI at three levels of analysis,
progressing from teams to organizations to society at large. At the team
level, this dissertation explores how organizations can effectively manage
human-AI teams to promote team collaboration. Based on interviews with
potential end users, a prototype team-AI collaboration system was developed
that allows human team members to individually configure AI agents by
assigning them different roles and personalities. This system was then
evaluated through a laboratory experiment in which human-AI teams collaborated
on a decision-making task. The results suggest that integrating configurable
AI team members into human teams can improve performance by introducing
complementary perspectives. However, human participants consistently favored
their own expertise for final team decisions, often disregarding superior
solutions provided by AI agents. Shifting the focus from collaboration
partners to environments, two studies investigate how the virtual reality
(VR)-based metaverse can facilitate team collaboration. In a laboratory
experiment, five teams performed a collaborative decision-making task using
either a VR-based metaverse platform (i.e., Meta Horizon Workrooms) or a
traditional videoconferencing tool (i.e., Zoom). The results indicate that
team collaboration in the metaverse can be a viable alternative to
videoconferencing tools, offering comparable (and in some areas superior)
levels of effectiveness, even in teams with minimal prior VR experience. At
the organizational level, this dissertation examines how organizations can
coordinate the learning activities of their human members and AI to enhance
overall organizational learning effectiveness. One study investigates the
mutual learning dynamics between humans and AI by introducing artificial
assistants (i.e., AI systems designed for recurring one-to-one collaboration)
that learn alongside humans. These artificial assistants can affirm or
challenge human knowledge while also contributing entirely new insights from
domains beyond their human partners’ expertise. Through a series of
agent-based simulations, the results show that artificial assistants can
reduce learning myopia, the human tendency to favor familiar strategies over
new and potentially better alternatives. Optimal outcomes occur when
organizations ensure that humans and AI are equally receptive to each other’s
insights, thus preventing an unbalanced learning process. A second study
examines how AI not only learns but also shapes organizational processes by
enacting its own beliefs. For example, AI can select job candidates based on
self-learned practices, gradually reshaping the organization’s view of what
makes a “good” candidate. Extending an established simulation model, the
results suggest that extensive coordination of enactment activities may be
unnecessary if humans and AI collaborate periodically to keep their beliefs
aligned. Together, these studies highlight that effective human-AI
collaboration depends on strategic managerial coordination to maximize
organizational learning and adaptability. At the societal level, this
dissertation explores strategies for integrating AI into society without
compromising cultural diversity. One study examines how different AI
integration strategies affect the evolution of cultural beliefs, using
agent-based simulations to model interactions between humans and AI. The
simulation results show that localized AI, designed to reflect regional or
national values, may inadvertently reduce cultural diversity by blending the
beliefs of neighboring social groups, challenging the assumption that
localization inherently preserves unique cultural identities. In contrast,
globalized AI, trained on data biased toward a dominant culture, may initially
support diversity but risks long-term polarization by pushing groups with
divergent beliefs toward (extreme) views that differ significantly from those
of the surrounding majority. These findings underscore that AI affects culture
in complex and sometimes unexpected ways, spreading beliefs while also
creating personalized echo chambers. To mitigate these risks, the simulation
results highlight the need for carefully designed policies that ensure AI
leaves space for different perspectives and does not unintentionally reinforce
social divides. The studies presented in this dissertation highlight that AI
is no longer merely a passive tool but an active participant in human learning
processes at the team, organizational, and societal levels. They demonstrate
that AI’s ability to both complement and challenge human expertise can enhance
collaboration, promote broader knowledge sharing, and mitigate human biases,
but only if its integration is carefully managed. Without deliberate
coordination, AI can instead reinforce inequalities, entrench dominant
narratives, and undermine diversity. This dissertation contributes to the
growing understanding of AI’s influence on human learning by offering
practical strategies for designing, integrating, and governing AI systems that
augment human capabilities. In doing so, it lays critical groundwork for
future research aimed at fostering human-AI collaborations that enhance human
learning and support the co-creation of knowledge without sacrificing unique
human knowledge and agency in the learning process. |