Kang, Xi. "Humans and Algorithms: Knowledge Workers Adaptation to Artificial Intelligence" (Job Market Paper)

• Nominee, Strategic Management Society Best Paper Prize (2020)

• Nominee, Strategic Management Society PhD Paper Prize (2020)

Self-adaptive artificial intelligence (AI) technologies can perform a growing number of cognitive tasks that used to depend on human judgement, raising debates about how AI may transform work, human capital and organization, especially in knowledge-intensive industries. This study highlights the important, yet often overlooked voluntary adaptation of knowledge workers to AI, and how it shapes the human–AI relationship through two mechanisms— a specialization mechanism and an information mechanism. Empirical analyses using proprietary data at the individual worker level and task level suggest that the adoption of AI significantly improves the performance of knowledge workers. AI provides implicit incentives for knowledge workers to exert effort and adapt, leading to a revised division of tasks between humans and AI.

Fan, Haoqing, Kang, Xi. “Performance Evaluations with Predictive Algorithms”

How do predictive algorithms affect the process of performance evaluation? Using a unique setting of the adoption of artificial intelligence in the evaluation process of mutual fund, we find that predictive algorithms can improve the quality of performance evaluations made by financial analysts by mitigating the impacts of conflicts of interest on security research. Exploited the staggered timing of adoption of the predictive algorithms, we find that analysts make less optimistic recommendations after algorithmic ratings are developed. This effect is particularly strong regarding the pillar rating devoted to evaluate fund managers’ personal style and expertise, which tend to be exceptionally positively biased. We further show that such improvement is greater for analysts who are socially connected to the fund managers being evaluated, because they are more likely to be subject to conflicts of interests.

Kang, Xi, Kim, Hyunjin. “Predictive Algorithms and Decision-making: Evidence from a Field Experiment”

Predictive algorithms have the potential to improve managerial decisions through information gains, although it may further exacerbate preexisting decision-making biases. Meanwhile, predictive algorithms may alter the role of decision-makers and change the decision-making processes and structures. This paper explores how predictive algorithms affect the way decision-makers formulate, reason, and cognitively approach decisions, and empirically evaluates the impacts on the decision-making quality and confidence. We conducted a series of field experiments across financial analysts in a global financial service company, wherein we investigated different ways of configurating predictive algorithms in the workflow of the decision-makers, and how they reason through and develop confidence in their decisions.

Kang, Xi, He, Xiaogang, Yu, Tieying, Cannella, Albert. “Are Family-Controlled Firms More Innovative? Family Control and Exploitative versus Exploratory Innovation”

Our study investigates the relationship between family control and innovation. By distinguishing between exploitative and exploratory innovations, we reconcile some conflicting theoretical arguments and empirical findings in the existing literature. Drawing insights from agency theory and research on family control, we argue that compared with nonfamily-controlled firms, familycontrolled firms are more likely to engage in exploitative innovation, and are less likely to engage in exploratory innovation. Additionally, increases in the degree of ownership-control separation strengthen the positive relationship between family control and exploitative innovation, while the level of firm-specific uncertainty strengthens the negative relationship between family control and exploratory innovation. Our hypotheses are empirically tested in China’s pharmaceutical industry from 2007 to 2016. The results offer strong support for our predictions.