Human values serve as guiding principles in people's lives. In the workplace, values shape priorities, influence decision-making, and predict commitment, performance, and satisfaction (Arieli, Sagiv, Roccas, 2021). Traditionally, values are measured through abstract self-reports, leaving unclear how they are enacted in real situations. Because values are embedded in language, interactions, and narratives, analyzing the way values appear in text offers a powerful window into how people construct meaning and enact their values in context.
Advances in natural language processing now enable the detection of “value instantiations,” the concrete ways people express values in real situations. In an ongoing project, expert annotators labeled employee-generated texts using Schwartz’s value system, creating ground truth for fine-tuning LLMs. These models reliably identify value expressions and culturally nuanced patterns that predict workplace outcomes better than traditional self-reports.
Building on this foundation, a second study investigates how values emerge in human-AI dialogue. In a controlled experiment, brief interactions with AI agents that embody openness to change or conservation values shifted participants’ linguistic patterns and influenced downstream behavior, such as creativity. These findings show that AI does not merely mirror human values but can actively prime and shape the expression of values.
Together, these projects demonstrate how integrating psychological theory with AI enables the generation of value-sensitive technologies that both detect and influence human values in context.