Reinforcement Learning for Education: Opportunities and Challenges

This survey article has grown out of the RL4ED workshop organized by the authors at the Educational Data Mining (EDM) 2021 conference. We organized this workshop as part of a community-building effort to bring together researchers and practitioners interested in the broad areas of reinforcement learning (RL) and education (ED). This article aims to provide an overview of the workshop activities and summarize the main research directions in the area of RL for ED.

Curriculum Design for Teaching via Demonstrations: Theory and Applications

We consider the problem of teaching via demonstrations in sequential decision-making settings. In particular, we study how to design a personalized curriculum over demonstrations to speed up the learner's convergence. We provide a unified curriculum strategy for two popular learner models: Maximum Causal Entropy Inverse Reinforcement Learning (MaxEnt-IRL) and Cross-Entropy Behavioral Cloning (CrossEnt-BC). Our unified strategy induces a ranking over demonstrations based on a notion of difficulty scores computed w.r.t. the teacher's optimal policy and the learner's current policy. Compared to the state of the art, our strategy doesn't require access to the learner's internal dynamics and still enjoys similar convergence guarantees under mild technical conditions. Furthermore, we adapt our curriculum strategy to teach a learner using domain knowledge in the form of task-specific difficulty scores when the teacher's optimal policy is unknown. Experiments on a car driving simulator environment and shortest path problems in a grid-world environment demonstrate the effectiveness of our proposed curriculum strategy.

Fair Clustering Using Antidote Data

Clustering algorithms are widely utilized for many modern data science applications. This motivates the need to make outputs of clustering algorithms fair. Traditionally, new fair algorithmic variants to clustering algorithms are developed for specific notions of fairness. However, depending on the application context, different definitions of fairness might need to be employed. As a result, new algorithms and analysis need to be proposed for each combination of clustering algorithm and fairness definition. Additionally, each new algorithm would need to be reimplemented for deployment in a real-world system. Hence, we propose an alternate approach to fairness in clustering where we augment the original dataset with a small number of data points, called antidote data. When clustering is undertaken on this new dataset, the output is fair, for the chosen clustering algorithm and fairness definition. We formulate this as a general bi-level optimization problem which can accommodate any center-based clustering algorithms and fairness notions. We then categorize approaches for solving this bi-level optimization for different problem settings. Extensive experiments on different clustering algorithms and fairness notions show that our algorithms can achieve desired levels of fairness on many real-world datasets with a very small percentage of antidote data added. We also find that our algorithms achieve lower fairness costs and competitive clustering performance compared to other state-of-the-art fair clustering algorithms.

Loss-Aversively Fair Classification

The use of algorithmic (learning-based) decision making in scenarios that affect human lives has motivated a number of recent studies to investigate such decision making systems for potential unfairness, such as discrimination against subjects based on their sensitive features like gender or race. However, when judging the fairness of a newly designed decision making system, these studies have overlooked an important influence on people's perceptions of fairness, which is how the new algorithm changes the status quo, i.e., decisions of the existing decision making system. Motivated by extensive literature in behavioral economics and behavioral psychology (prospect theory), we propose a notion of fair updates that we refer to as loss-averse updates. Loss-averse updates constrain the updates to yield improved (more beneficial) outcomes to subjects compared to the status quo. We propose tractable proxy measures that would allow this notion to be incorporated in the training of a variety of linear and non-linear classifiers. We show how our proxy measures can be combined with existing measures for training nondiscriminatory classifiers. Our evaluation using synthetic and real-world datasets demonstrates that the proposed proxy measures are effective for their desired tasks.

Reward Poisoning in Reinforcement Learning: Attacks Against Unknown Learners in Unknown Environments

We study black-box reward poisoning attacks against reinforcement learning (RL), in which an adversary aims to manipulate the rewards to mislead a sequence of RL agents with unknown algorithms to learn a nefarious policy in an environment unknown to the adversary a priori. That is, our attack makes minimum assumptions on the prior knowledge of the adversary: it has no initial knowledge of the environment or the learner, and neither does it observe the learner's internal mechanism except for its performed actions. We design a novel black-box attack, U2, that can provably achieve a near-matching performance to the state-of-the-art white-box attack, demonstrating the feasibility of reward poisoning even in the most challenging black-box setting.

Defense Against Reward Poisoning Attacks in Reinforcement Learning

We study defense strategies against reward poisoning attacks in reinforcement learning. As a threat model, we consider attacks that minimally alter rewards to make the attacker's target policy uniquely optimal under the poisoned rewards, with the optimality gap specified by an attack parameter. Our goal is to design agents that are robust against such attacks in terms of the worst-case utility w.r.t. the true, unpoisoned, rewards while computing their policies under the poisoned rewards. We propose an optimization framework for deriving optimal defense policies, both when the attack parameter is known and unknown. Moreover, we show that defense policies that are solutions to the proposed optimization problems have provable performance guarantees. In particular, we provide the following bounds with respect to the true, unpoisoned, rewards: a) lower bounds on the expected return of the defense policies, and b) upper bounds on how suboptimal these defense policies are compared to the attacker's target policy. We conclude the paper by illustrating the intuitions behind our formal results, and showing that the derived bounds are non-trivial.

Policy Teaching in Reinforcement Learning via Environment Poisoning Attacks

We study a security threat to reinforcement learning where an attacker poisons the learning environment to force the agent into executing a target policy chosen by the attacker. As a victim, we consider RL agents whose objective is to find a policy that maximizes reward in infinite-horizon problem settings. The attacker can manipulate the rewards and the transition dynamics in the learning environment at training-time, and is interested in doing so in a stealthy manner. We propose an optimization framework for finding an optimal stealthy attack for different measures of attack cost. We provide lower/upper bounds on the attack cost, and instantiate our attacks in two settings: (i) an offline setting where the agent is doing planning in the poisoned environment, and (ii) an online setting where the agent is learning a policy with poisoned feedback. Our results show that the attacker can easily succeed in teaching any target policy to the victim under mild conditions and highlight a significant security threat to reinforcement learning agents in practice.

The Teaching Dimension of Kernel Perceptron

Algorithmic machine teaching has been studied under the linear setting where exact teaching is possible. However, little is known for teaching nonlinear learners. Here, we establish the sample complexity of teaching, aka teaching dimension, for kernelized perceptrons for different families of feature maps. As a warm-up, we show that the teaching complexity is $\Theta(d)$ for the exact teaching of linear perceptrons in $\mathbb{R}^d$, and $\Theta(d^k)$ for kernel perceptron with a polynomial kernel of order $k$. Furthermore, under certain smooth assumptions on the data distribution, we establish a rigorous bound on the complexity for approximately teaching a Gaussian kernel perceptron. We provide numerical examples of the optimal (approximate) teaching set under several canonical settings for linear, polynomial and Gaussian kernel perceptrons.

Preference-Based Batch and Sequential Teaching

Algorithmic machine teaching studies the interaction between a teacher and a learner where the teacher selects labeled examples aiming at teaching a target hypothesis. In a quest to lower teaching complexity, several teaching models and complexity measures have been proposed for both the batch settings (e.g., worst-case, recursive, preference-based, and non-clashing models) and the sequential settings (e.g., local preference-based model). To better understand the connections between these models, we develop a novel framework that captures the teaching process via preference functions $\Sigma$. In our framework, each function $\sigma \in \Sigma$ induces a teacher-learner pair with teaching complexity as $TD(\sigma)$. We show that the above-mentioned teaching models are equivalent to specific types/families of preference functions. We analyze several properties of the teaching complexity parameter $TD(\sigma)$ associated with different families of the preference functions, e.g., comparison to the VC dimension of the hypothesis class and additivity/sub-additivity of $TD(\sigma)$ over disjoint domains. Finally, we identify preference functions inducing a novel family of sequential models with teaching complexity linear in the VC dimension: this is in contrast to the best-known complexity result for the batch models, which is quadratic in the VC dimension.

Synthesizing Tasks for Block-based Programming

Block-based visual programming environments play a critical role in introducing computing concepts to K-12 students. One of the key pedagogical challenges in these environments is in designing new practice tasks for a student that match a desired level of difficulty and exercise specific programming concepts. In this paper, we formalize the problem of synthesizing visual programming tasks. In particular, given a reference visual task $\rm T^{in}$ and its solution code $\rm C^{in}$, we propose a novel methodology to automatically generate a set $\{(\rm T^{out}, \rm C^{out})\}$ of new tasks along with solution codes such that tasks $\rm T^{in}$ and $\rm T^{out}$ are conceptually similar but visually dissimilar. Our methodology is based on the realization that the mapping from the space of visual tasks to their solution codes is highly discontinuous; hence, directly mutating reference task $\rm T^{in}$ to generate new tasks is futile. Our task synthesis algorithm operates by first mutating code $\rm C^{in}$ to obtain a set of codes $\{\rm C^{out}\}$. Then, the algorithm performs symbolic execution over a code $\rm C^{out}$ to obtain a visual task $\rm T^{out}$; this step uses the Monte Carlo Tree Search (MCTS) procedure to guide the search in the symbolic tree. We demonstrate the effectiveness of our algorithm through an extensive empirical evaluation and user study on reference tasks taken from the \emph{Hour of the Code: Classic Maze} challenge by \emph{Code.org} and the \emph{Intro to Programming with Karel} course by \emph{CodeHS.com}.