An official website of the United States government
Standard methods for synthesis of control policies in Markov decision processes with unknown transition probabilities largely rely on a combination of exploration and exploitation. While these methods often offer theoretical guarantees on system performance, the number of time steps and samples needed to initially explore the environment before synthesizing a well-performing control policy is impractically large. This paper partially alleviates such a burden by incorporating a priori existing knowledge into learning, when such knowledge is available. Based on prior information about bounds on the differences between the transition probabilities at different states, we propose a learning approach where the transition probabilities at a given state are not only learned from outcomes of repeatedly performing a certain action at that state, but also from outcomes of performing actions at states that are known to have similar transition probabilities. Since the directly obtained information is more reliable at determining transition probabilities than second-hand information, i.e., information obtained from similar but potentially slightly different states, samples obtained indirectly are weighted with respect to the known bounds on the differences of transition probabilities. While the proposed strategy can naturally lead to errors in learned transition probabilities, we show that, by proper choice of the weights, such errors can be reduced, and the number of steps needed to form a near-optimal control policy in the Bayesian sense can be significantly decreased.
more »
« less
When Less Is More: Experimental Evidence on Information Delivery During India’s Demonetisation
Abstract In disseminating information, policymakers face a choice between broadcasting to everyone and informing a small number of “seeds” who then spread the message. While broadcasting maximises the initial reach of messages, we offer theoretical and experimental evidence that it need not be the best strategy. In a field experiment during the 2016 Indian demonetisation, we delivered policy information, varying three dimensions of the delivery method at the village level: initial reach (broadcasting versus seeding); whether or not we induced common knowledge of who was initially informed; and number of facts delivered. We measured three outcomes: the volume of conversations about demonetisation, knowledge of demonetisation rules, and choice quality in a strongly incentivised policy-dependent decision. On all three outcomes, under common knowledge, seeding dominates broadcasting; moreover, adding common knowledge makes seeding more effective but broadcasting less so. We interpret our results via a model of image concerns deterring engagement in social learning, and we support this interpretation with evidence on differential behaviour across ability categories.
more »
« less
- Award ID(s):
- 1658940
- PAR ID:
- 10520966
- Publisher / Repository:
- Oxford University Press
- Date Published:
- Journal Name:
- Review of Economic Studies
- Volume:
- 91
- Issue:
- 4
- ISSN:
- 0034-6527
- Format(s):
- Medium: X Size: p. 1884-1922
- Size(s):
- p. 1884-1922
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract We evaluate the cost‐effectiveness of two early childhood interventions that use instructional coaching and parent coaching as levers for improvement. The study design allows us to compare the individual effects of each intervention as well as their combined effect on student outcomes. We find that teachers receiving instructional coaching improve their use of evidence‐based instructional practices, while families receiving parent coaching show increases in numerous responsive parenting behaviors associated with positive child outcomes. Both interventions demonstrate positive impacts on students, but only parent coaching shows statistically significant effects across a range of student outcomes. Instructional coaching alone is substantially less costly and may therefore be the most cost‐effective of the three treatment conditions; however, small sample sizes limit our ability to reach definitive conclusions. Policy simulations suggest that implementing these interventions could raise the overall cost‐effectiveness of Head Start by at least 16 percent.more » « less
-
Abstract Here, we systematically review research on teaching knowledge in the context of undergraduate STEM education, with particular attention to what this research reveals about knowledge that is important for evidence-based teaching. Evidence-based teaching can improve student outcomes in undergraduate STEM education. However, the enactment of promising evidence-based teaching strategies depends greatly on the instructor and potentially on the teaching knowledge they are able to deploy. The review includes an overview of prevalent teaching knowledge theory, including pedagogical content knowledge, mathematical knowledge for teaching, and pedagogical knowledge. We compare and contrast teaching knowledge theory and terminology across STEM disciplines in order to build bridges for researchers across disciplines. Our search for peer-reviewed investigations of teaching knowledge in undergraduate science, engineering and mathematics yielded 45 papers. We examined the theoretical frameworks used in each study and analyzed study approaches, comparing across disciplines. Importantly, we also synthesized findings from research conducted in the context of evidence-based teaching. Overall, teaching knowledge research is sparse and siloed by discipline, and we call for collaborative work and better bridge-building across STEM disciplines. Though disciplinary divergences are common in discipline-based education research, the effect is magnified in this research area because the theoretical frameworks are themselves siloed by discipline. Investigations of declarative knowledge were common, and we call for increased attention to knowledge used in the practice of teaching. Finally, there are not many studies examining teaching knowledge in the context of evidence-based teaching, but the existing work suggests that components of pedagogical content knowledge, pedagogical knowledge, and content knowledge influence the implementation of evidence-based teaching. We describe implications for future teaching knowledge research. We also call on those who develop and test evidence-based strategies and curriculum to consider, from the beginning, the teaching knowledge needed for effective implementation.more » « less
-
The First2 Network is an alliance of higher education institutions across the State of West Virginia striving to improve science, technology, engineering or math (STEM) education by supporting rural, first-generation, and underrepresented college students pursuing STEM majors. Over the summers of 2019 and 2020, the First2 Network delivered two-week summer research immersion experiences at various institutions throughout West Virginia. The 2019 program was delivered on-campus at four universities while the 2020 program was delivered virtually, due to the COVID-19 pandemic, across nine sites. Before and after the immersion experience, students who participated in the program completed a variety of survey questionnaires for the assessment of their interests, expectations, identity and belonging in STEM. We found that the in-person experience in 2019 had better outcomes compared to the virtual experience, suggesting students conducting research directly under their faculty supervisors in-person and on-site will have a more positive impact on their STEM education and career. However, participation in the virtual format still resulted in an improvement in belonging and STEM identity, indicating that connecting with students remotely is still worthwhile when it is the most viable option. The student population in Appalachia faces a number of academic barriers, so there is much to gain by finding new ways to reach as many students as possible with early career development programs.more » « less
-
This report documents the program and the outcomes of Dagstuhl Seminar 23031 Frontiers of Information Access Experimentation for Research and Education, which brought together 38 participants from 12 countries. The seminar addressed technology-enhanced information access (information retrieval, recommender systems, natural language processing) and specifically focused on developing more responsible experimental practices leading to more valid results, both for research as well as for scientific education. The seminar featured a series of long and short talks delivered by participants, who helped in setting a common ground and in letting emerge topics of interest to be explored as the main output of the seminar. This led to the definition of five groups which investigated challenges, opportunities, and next steps in the following areas:reality check, i.e. conducting real-world studies, human-machine-collaborative relevance judgment frameworks, overcoming methodological challenges in information retrieval and recommender systems through awareness and education, results-blind reviewing, and guidance for authors. Date:15--20 January 2023. Website:https://www.dagstuhl.de/23031.more » « less
