Search

Machine learning (ML) methods are proliferating in scientific research. However, the adoption of these methods has been accompanied by failures of validity, reproducibility, and generalizability. These failures can hinder scientific progress, lead to false consensus around invalid claims, and undermine the credibility of ML-based science. ML methods are often applied and fail in similar ways across disciplines. Motivated by this observation, our goal is to provide clear recommendations for conducting and reporting ML-based science. Drawing from an extensive review of past literature, we present the REFORMS checklist (recommendations for machine-learning-based science). It consists of 32 questions and a paired set of guidelines. REFORMS was developed on the basis of a consensus of 19 researchers across computer science, data science, mathematics, social sciences, and biomedical sciences. REFORMS can serve as a resource for researchers when designing and implementing a study, for referees when reviewing papers, and for journals when enforcing standards for transparency and reproducibility.

Moral behavior is more prevalent when individuals cannot easily distort their beliefs self-servingly. Do individuals seek to limit or enable their ability to distort beliefs? How do these choices affect behavior? Experiments with over 9,000 participants show preferences are heterogeneous—30 percent of participants prefer to limit belief distortion, while over 40 percent prefer to enable it, even if costly. A random assignment mechanism reveals that being assigned to the preferred environment is necessary for curbing or enabling self-serving behavior. Third parties can anticipate these effects, suggesting some sophistication about the cognitive constraints to belief distortion. (JEL C91, D82, D83, D91)

Willingness to vaccinate and test are critical in the COVID-19 pandemic. We study the effects of two measures to increase the support of vaccination and testing: defaults and monetary compensations. Some organizations, such as restaurants, fire departments, hospitals, or governments in some countries, have used these measures. Yet there is the concern that compensations could erode intrinsic motivation and decrease vaccination intentions. We show that, in the early stages of the pandemic, both approaches, compensations and defaults, significantly increased COVID-19 test demand and vaccine intentions. Compensations for vaccines, however, need to be large enough because low compensations can backfire. We estimate heterogeneous treatment effects to document which groups are more likely to respond to these measures. The results show that defaults and avoidance of small compensations are especially important for individuals who are more skeptical of the vaccine, measured by their trust in the vaccine and their political views. Hence, both measures could be used in a targeted manner to achieve stronger results. This paper was accepted by Yan Chen, behavioral economics and decision analysis. Supplemental Material: The online appendix and data are available at https://doi.org/10.1287/mnsc.2022.4405 .

Ignorance enables individuals to act immorally. This is well known in policy circles, in which there is keen interest in lowering moral ignorance. In this paper, we study how the demand for moral ignorance responds to monetary incentives and how the demand curve for ignorance reacts to social norm messages. We propose a simple behavioral model in which individuals suffer moral costs when behaving selfishly in the face of moral information. In several experiments, we find that moral ignorance decreases by more than 30 percentage points with small monetary incentives, but we find no significant change with social norm messages, and we document strong persistence of ignorance across moral contexts. Our findings indicate that rather simple messaging interventions may have limited effects on ignorance. In contrast, changes in incentives could be highly effective. This paper was accepted by Yan Chen, behavioral economics and decision analysis.

AbstractRisk attitudes determine decision making by individuals. Theoretically, heterogeneity in risk attitudes also affects household behaviour, increasing or decreasing marital instability, depending on whether households benefit primarily from sharing public goods or risk. Using a nationally representative sample with repeated measurements of risk attitudes, this paper shows that intrahousehold heterogeneity in risk attitudes is an economically and statistically significant predictor of future marital instability. Consistent with models of public good sharing, preference heterogeneity is associated with less public good sharing. Furthermore, risk attitudes converge within stable households over time. These findings highlight the empirical relevance of risk attitudes in explaining household behaviour.

University plays a vital role in the social development of students, and, in line with this premise, the main foundations for the creation of the European Higher Education Area are based on the social perspective for its promotion. The introduction of the Spanish university system into the European Higher Education Area has had an impact on initial teacher training. In order to maintain a balance from the social perspective, this work analyses the student's perception of a bachelor's degree in early childhood education and primary education and a master's in secondary education on the implications of the European convergence process for their initial training. It is a descriptive study that uses an "ad hoc" scale as an instrument for data collection, addressed to students from two Spanish universities (n=682). Results show the positive evaluation of the student towards the practicum, understood as a complement to their academic training, as well as the coherence between the theoretical and practical contents in the different subjects, among other aspects. However, it also criticizes the development of subjects taught by several professors, as well as the lack of information disseminated by the university on the implications that this European process has generated in the teaching-learning process. These perceptions are more favorable in master's students than in undergraduate students. This allows the identification of the potentialities and limitations of the European process for the initial teacher training.

The current offer of training courses for university teachers is due, among other needs, to the implementation of an educational model based on student learning, promoting the use of active methodologies for their motivation and academic performance. An exploratory-descriptive and ideographic study is presented where the main technique is the analysis of content. To this end, 15 Spanish universities with the greatest prestige were analyzed according to the parameters of the Academic Ranking of World Universities (ARWU) in relation to the topics related to active methodologies and the profile of university teachers. The results show the different teacher training courses that have been carried out during the 2019/2020 academic year. We find that there are more universities that present more training of this type than others, and who this type of training is aimed at: new teachers and teachers with professional experience. The conclusions are related to the importance of the courses for the professional development of university teachers, since they should not be anchored in the same methodology, but should be open to new challenges and always taking into account the students, enhancing their motivation and academic performance.

Mistakes and overconfidence in detecting lies could help lies spread. Participants in our experiments observe videos in which senders either tell the truth or lie, and are incentivized to distinguish between them. We find that participants fail to detect lies, but are overconfident about their ability to do so. We use these findings to study the determinants of sharing and its effect on lie detection, finding that even when incentivized to share truthful videos, participants are more likely to share lies. Moreover, the receivers are more likely to believe shared videos. Combined, the tendency to believe lies increases with sharing. (JEL C91, D83, D91, L82)

What is the value of pledges if they are often reneged upon? In this paper, we show—both theoretically and experimentally—that pledges can be used to screen donors and to better understand their motives for giving. In return, nonprofit managers can use the information they glean from pledges to better target future charitable giving appeals and interventions to donors, such as expressions of gratitude. In an experiment, we find that offering the option to pledge gifts induces self-selection. If expressions of gratitude are then targeted to individuals who select into pledges, reneging can be significantly reduced. Our findings provide an explanation for the potential usefulness of pledges. This paper was accepted by Yan Chen, decision analysis.