Research

AI and Decision Making

The newest research conducted by Professor Li examines the role AI can play in human decision making. New projects in this general area include:

1. Human-AI collaboration can benefit from complementarity but do people realize? We predict people prefer to work with others that perform like themselves and thus exhibit complementary neglect. Participants chose between two partners: Supporter A performed like a human and thus got the same trials right/wrong as participants. Supporter B performed like an AI and thus was right on trials that participants got wrong, but was wrong for trials that participants got right. Although Supporter B was more complementary, people preferred Supporter A. People who chose Supporter B performed better due to complementarity.

2. The use of artificial intelligence (AI) to generate text has become increasingly common, especially with the announcement of ChatGPT in December 2022. This project aims to explore whether people are more persuaded by text written by other humans or by AI-generated text.

3. People often compliment each other in organizational settings, whether a manager, fellow team member, or a subordinate. How good are people at writing such workplace compliments? We use Natural Language Processing (NLP) algorithms to examine hundreds of human-written compliments that have been rated on how effective they are at making the receiver of the compliment a) feel good, b) want to work together in the future, and c) be friends. We then examine how well people tasked with writing for one of these three goals are at writing compliments and machine learning to help write better compliments.

Time Preference

Many decisions require intertemporal tradeoffs—from the mundane (i.e., whether to order dessert or floss tonight) to the consequential (i.e., whether to start smoking or how much to save for retirement). One of Professor Li's primary streams of research seeks to understand when and why people behave more or less patiently when faced with these "intertemporal choices."

One set of papers has examined the impact of emotions on time preferences. One paper (2013 Psychological Science) examined whether the devaluation of self that accompanies feelings of sadness can lead to an increased urgency to acquire things. We tested this by combining mood induction procedures with intertemporal choices. We found that sadness impacted cognitions by biasing thoughts towards ones that supported impatient choices. We also found that sadness preferentially made people more impatient when immediate outcomes were available but was less impactful when the earliest possible outcome was at least 2 weeks away—i.e., it ramped up instant gratification. In a follow-up paper (2014 Psychological Science), we found that participants made to feel happier did not choose more patiently. Instead, a different positive emotion, gratitude, substantially increased patience in financial choices. This research was theoretically important as the first to find that a specific emotion can reduce impatience, whereas many theories suggest that emotions are detrimental to rational choice. Professor Li also wrote a review paper on “Emotion and Decision Making” in Annual Review of Psychology, organizing 35 years of research on emotion and decision making into eight major themes.

While emotions have short term influences on intertemporal choices, longer term influences also matter, such as cognitive aging. Both lay theory and past research suggest two conflicting views about aging: that age brings wisdom, and that age brings diminished cognitive ability. It turns out there is truth to both views: fluid intelligence—i.e., the ability to generate, transform and manipulate information—declines with age, but is it possible that older adults’ greater crystallized intelligence—their knowledge, expertise, and life experience—can help offset declining fluid intelligence in terms of making good decisions? In one paper (2013 Psychology and Aging), Professor Li found that greater experience and acquired knowledge from a lifetime of decision-making may provide older people with another way to make good decisions. A follow-up paper (2015 Proceedings of the National Academy of Sciences) extended this work by examining the relationship between cognitive aging and real-world financial decisions.

Finally, two papers advance the measurement of time preferences. First, in a Journal of Marketing Research (2022) paper, Professor Li examined the elicitation methods commonly used to measure time preferences and other individual differences by marketers, managers, and policy makers. To predict real-world behaviors, we often ask people questions about their preferences. But each question changes how people think: They learn to use a specific strategy for the task, which may not match how they think in real life. This can lower the accuracy of our predictions. We found that people used more task-specific strategies as they answered more questions, and this hurt the accuracy of predicting other tasks and real behaviors. We also found that the best accuracy was achieved after less than seven questions for both types of tasks. So, when asking questions to measure preferences, more is not better. Second, a paper published in Journal of Experimental Psychology: General (2023) conducted a comprehensive investigation of how well time preference predicts behavior—1) examining more behaviors, 2) controlling for more covariates, and 3) using a test-retest design to account for measurement error. We found correlations that are mostly modest and highly variable across behaviors. We also found that time preference researchers (N = 55) tended to overestimate the predictive power of time preference estimates.

Incidental Role of Time in Decision Making

A second stream of research examines the influence that time has on decision making even when it is incidental or peripheral.

Professor Li and BEDLab alumnus, Yun Jie, (Journal of Retailing 2022) studied the role of temporal cues on consumer decisions. Previous research has looked at how people like new options that are unique, original, or novel. We focus on how people like new options that are only different in time (e.g., when they were made, sold, or bought). We ask if people prefer newer options, even if they are not better in any other way. We find that people do prefer and pay more for “merely” newer options because they think “newer” means “better.” We show this for choices of headphones, posters, and lottery tickets with different time tags that do not change their quality.

The temporal cues paper studies how explicit time tags affect choices. But time also passes implicitly, and this can change how we choose. A paper in Journal of Behavioral Decision Making (2009) showed that people are biased by how vivid recent experiences are. When choosing from good options, the last one seems best because it is fresher in memory. When choosing from bad options, the last one seems worst, but earlier ones seem better than they really are.

Finally, although global warming is a phenomenon that can only be detected over long periods of time, a Psychological Science (2011) paper found that people’s judgments about global warming are influenced by temperature over much shorter time horizons. People who felt that today was hotter than usual were more likely to believe in and donate to global warming causes, in both the US and Australia, and in both winter and summer. A follow-up in Current Opinion in Behavioral Sciences (2021) reviewed a decade of studies replicating and extending the original finding and conducted a formal meta-analysis on the “local warming” phenomenon. We found that this effect seems to be robust, although with considerable heterogeneity in effect size.