More Like this

1. Hierarchical Reinforcement Learning for Pedagogical Policy Induction

   Zhou, G. ( January 2019 , In Proceedings of the 19th International Conference on Artificial Intelligence in Education (AIED-2018))

   In interactive e-learning environments such as Intelligent Tutoring Systems, there are pedagogical decisions to make at two main levels of granularity: whole problems and single steps. Recent years have seen growing interest in data-driven techniques for such pedagogical decision making, which can dynamically tailor students’ learning experiences. Most existing data-driven approaches, however, treat these pedagogical decisions equally, or independently, disregarding the long-term impact that tutor decisions may have across these two levels of granularity. In this paper, we propose and apply an offline, off-policy Gaussian Processes based Hierarchical Reinforcement Learning (HRL) framework to induce a hierarchical pedagogical policy that makes decisions at both problem and step levels. In an empirical classroom study with 180 students, our results show that the HRL policy is significantly more effective than a Deep Q-Network (DQN) induced policy and a random yet reasonable baseline policy. 

   more » « less
2. Hierarchical Reinforcement Learning for Pedagogical Policy Induction

   Guojing Zhou, Hamoon Azizsoltani ( January 2019 , International Conference on Artificial Intelligence in Education)

   Abstract.In interactive e-learning environments such as Intelligent Tutor-ing Systems, there are pedagogical decisions to make at two main levels of granularity: whole problems and single steps. Recent years have seen grow-ing interest in data-driven techniques for such pedagogical decision making, which can dynamically tailor students’ learning experiences. Most existing data-driven approaches, however, treat these pedagogical decisions equally, or independently, disregarding the long-term impact that tutor decisions may have across these two levels of granularity. In this paper, we propose and apply an offline, off-policy Gaussian Processes based Hierarchical ReinforcementLearning (HRL) framework to induce a hierarchical pedagogical policy that makes decisions at both problem and step levels. In an empirical classroom study with 180 students, our results show that the HRL policy is significantly more effective than a Deep Q-Network (DQN) induced policy and a random yet reasonable baseline policy. 

   more » « less
3. Long-Term Impacts of Fair Machine Learning

   https://doi.org/10.1177/1064804619884160  

   Zhang, Xueru ; Khalili, Mohammad_Mahdi ; Liu, Mingyan ( October 2019 , Ergonomics in Design: The Quarterly of Human Factors Applications)

   Machine learning models developed from real-world data can inherit potential, preexisting bias in the dataset. When these models are used to inform decisions involving human beings, fairness concerns inevitably arise. Imposing certain fairness constraints in the training of models can be effective only if appropriate criteria are applied. However, a fairness criterion can be defined/assessed only when the interaction between the decisions and the underlying population is well understood. We introduce two feedback models describing how people react when receiving machine-aided decisions and illustrate that some commonly used fairness criteria can end with undesirable consequences while reinforcing discrimination.

    

   more » « less
4. Applying the CAPE framework to measure equity and inform policy in computer science education

   https://doi.org/10.1177/14782103221074467  

   Warner, Jayce R ; Fletcher, Carol L ; Martin, Nicole D ; Baker, Stephanie N ( January 2022 , Policy Futures in Education)

   The push to make computer science (CS) education available to all students has been closely followed by increased efforts to collect and report better data on where CS is offered, who is teaching CS, and which students have access to, enroll in, and ultimately benefit from learning CS. These efforts can be highly influential on the evolution of CS education policy, as education leaders and policymakers often rely heavily on data to make decisions. Because of this, it is critical that CS education researchers understand how to collect, analyze, and report data in ways that reflect reality without masking disparities between subpopulations. Similarly, it is important that CS education leaders and policymakers understand how to judiciously interpret the data and translate information into action to scale CS education in ways designed to eliminate inequities. To that end, this article expands on recent research regarding the use of data to assess and inform progress in scaling and broadening participation in CS education. We describe the CAPE framework for assessing equity with respect to the capacity for, access to, participation in, and experience of CS education and explicate how it can be applied to analyze and interpret data to inform policy decisions at multiple levels of educational systems. We provide examples using large, statewide datasets containing educational and demographic information for K-12 students and schools, thereby giving leaders and policymakers a roadmap to assess and address issues of equity in their own schools, districts, or states. We compare and contrast different approaches to measuring and reporting inequities and discuss how data can influence the future of CS education through its impact on policy. 

   more » « less
5. Hierarchical Reinforcement Learning for Pedagogical Policy Induction.

   https://doi.org/10.1007/978-3-030-23204-7_45  

   Zhou, G. ; Azizsoltani, H. ; Ausin, M.S. ; Barnes, T. ; Chi, M. ( January 2020 , n Proceedings of the 29th International Joint Conference on Artificial Intelligence (IJCAI-2020))

   null (Ed.)

   Abstract: In interactive e-learning environments such as Intelligent Tutoring Systems, there are pedagogical decisions to make at two main levels of granularity: whole problems and single steps. In recent years, there is growing interest in applying datadriven techniques for adaptive decision making that can dynamically tailor students’ learning experiences. Most existing data-driven approaches, however, treat these pedagogical decisions equally, or independently, disregarding the long-term impact that tutor decisions may have across these two levels of granularity. In this paper, we propose and apply an offline Gaussian Processes based Hierarchical Reinforcement Learning (HRL) framework to induce a hierarchical pedagogical policy that makes decisions at both problem and step levels. An empirical classroom study shows that the HRL policy is significantly more effective than a Deep QNetwork (DQN) induced policy and a random yet reasonable baseline policy. 

   more » « less