Adam Eck

Lessons Learned from Comprehensive Deployments of Multiagent CSCL Applications I-MINDS and ClassroomWiki

Recent years have seen a surge in the use of intelligent computer-supported collaborative learning (CSCL) tools for improving student learning in traditional classrooms. However, adopting such a CSCL tool in a classroom still requires the teacher to develop (or decide on which to adopt) the CSCL tool and the CSCL script, design the relevant pedagogical aspects (i.e., the learning objectives, assessment method, etc.) to overcome the associated challenges (e.g., free riding, student assessment, forming student groups that improve student learning, etc). We have used a multiagent-based system to develop a CSCL application and multiagent-frameworks to form student groups that improve student collaborative learning. In this paper, we describe the contexts of our three generations of CSCL applications (i.e., I-MINDS and ClassroomWiki) and provide a set of lessons learned from our deployments in terms of the script, tool, and pedagogical aspects of using CSCL. We believe that our lessons would allow 1) the instructors and students to use intelligent CSCL applications more effectively and efficiently, and help to improve the design of such systems, and 2) the researchers to gain additional insights into the impact of collaborative learning theories when they are applied to real-world classrooms.

Observer effect from stateful resources in agent sensing

In many real-world applications of multi-agent systems, agent reasoning suffers from bounded rationality caused by both limited resources and limited knowledge. When agent sensing to overcome its knowledge limitations also requires resource use, the agent’s knowledge refinement is affected due to its inability to always sense when and as accurately as needed, further leading to poor decision making. In this paper, we consider what happens when sensing actions require the use of stateful resources, which we define as resources whose state-dependent behavior changes over time based on usage. Current literature addressing agent sensing with limited resources primarily investigates stateless resources, such as avoiding the use of too much time or energy during sensing. However, sensing itself can change the state of a resource, and thus its behavior, which affects both the information gathered and the resulting knowledge refinement. This produces a phenomenon where the sensing action can and will distort its own outcome (and potentially future outcomes), termed the Observer Effect (OE) after the similar phenomenon in the physical sciences. Under this effect, when deliberating about when and how to perform sensing that requires use of stateful resources, an agent faces a strategic tradeoff between satisfying the need for (1) knowledge refinement to support its reasoning, and (2) avoiding knowledge corruption due to distorted sensing outcomes. To address this tradeoff, we model sensing action selection as a partially observable Markov decision process where an agent optimizes knowledge refinement while considering the (possibly hidden) state of the resources used during sensing. In this model, the agent uses reinforcement learning to learn a controller for action selection, as well as how to predict expected knowledge refinement based on resource use during sensing. Our approach is unique from other bounded rationality and sensing research as we consider how to make decisions about sensing with stateful resources that produce side effects such as the OE, as opposed to simply using stateless resources with no such side effect. We evaluate our approach in a fully and partially observable agent mining simulation. The results demonstrate that considering resource state and the OE during sensing action selection through our approach (1) yielded better knowledge refinement, (2) appropriately balanced current and future refinement to avoid knowledge corruption, and (3) exploited the relationship (i.e., high, positive correlation) between sensing and task performance to boost task performance through improved sensing. Further, our methodology also achieved good knowledge refinement even when the OE is not present, indicating that it can improve sensing performance in a wide variety of environments. Finally, our results also provide insights into the types and configurations of learning algorithms useful for learning within our methodology.

Potential-based reward shaping for finite horizon online POMDP planning

In this paper, we address the problem of suboptimal behavior during online partially observable Markov decision process (POMDP) planning caused by time constraints on planning. Taking inspiration from the related field of reinforcement learning (RL), our solution is to shape the agent’s reward function in order to lead the agent to large future rewards without having to spend as much time explicitly estimating cumulative future rewards, enabling the agent to save time to improve the breadth planning and build higher quality plans. Specifically, we extend potential-based reward shaping (PBRS) from RL to online POMDP planning. In our extension, information about belief states is added to the function optimized by the agent during planning. This information provides hints of where the agent might find high future rewards beyond its planning horizon, and thus achieve greater cumulative rewards. We develop novel potential functions measuring information useful to agent metareasoning in POMDPs (reflecting on agent knowledge and/or histories of experience with the environment), theoretically prove several important properties and benefits of using PBRS for online POMDP planning, and empirically demonstrate these results in a range of classic benchmark POMDP planning problems.

Testing collaborative traffic over wireless protocols

As technology advances, new opportunities present themselves for computer supported learning in educational environments. One of the enabling technologies for these opportunities is wireless networking. Without being tethered to wires, students can use laptops and other wireless devices both in and out of classrooms for group work and self-paced learning. This will help prepare them for a collaboration-driven global workplace, and it promotes new opportunities for distance education. However, wireless communications currently suffer from several inhibiting problems stemming from the underlying network protocol used by the software. The Transmission Control Protocol (TCP) is the most utilized standard in networking and is well known for its reliability, but it was built to work with wired networks and its assumptions for the cause of packet loss do not fit with the reality of wireless environments: loss is caused by inherent problems in the network layer and not data collisions. Because of this, the user datagram protocol (UDP) is often used in wireless environments since it can achieve higher levels of throughput by not slowing down traffic to prevent data collisions, unlike TCP, and it supports multicast transmissions. However, UDP provides no reliability; if any packet is lost, it is gone forever. In educational environments, packet loss is not acceptable because every piece of information sent by a teacher or student is important, and throughput is necessary because slow software inhibits productivity and learning. To help resolve this problem, we have tested the experimental pragmatic general multicast protocol (PGM), which supports both reliability through lost packet retrieval and high throughput by not scaling traffic and by using multicast transmissions. Our tests in a real-time collaborative classroom environment of over thirty wireless laptops via the Microsoft ConferenceXP platform have promising results: we encountered no problems in throughput or packet loss. However, we discovered problems with message latency, so we conducted further tests and present our results in this paper.

Investigating Differences in Wiki-based Collaborative Activities between Student Engagement Profiles in CS1

Introductory computer science courses are being increasingly taught using technology-mediated instruction and e-learning environments. The software and technology in such courses could benefit from the use of student models to inform and guide customized support tailored to the needs of individual students. In this paper, we investigate how student motivated engagement profiles developed in educational research can be used as such models to predict student behaviors. These models are advantageous over those learned directly from observing individual students, as they rely on different data that can be available a priori before students use the technology. Using tracked behaviors of 249 students from 7 CS1 courses over the span of 3 semesters, we discover that students with different engagement profiles indeed behave differently in an online, wiki-based CSCL system while performing collaborative creative thinking exercises, and the differences between students are primarily as expected based on the differences in the profiles. Thus, such profiles could be useful as student models for providing customized support in e-learning environments in CS1 courses.

MineralMiner: An active sensing simulation environment

One important problem in multiagent systems is determining how to gather information through sensing to support agent reasoning. This problem commonly arises in real-world applications such as robotics, mixed initiative systems, and others. One promising solution is to use active sensing to explicitly reason about the benefits e.g., information gain, accuracy and costs e.g., resource use, knowledge corruption of sensing actions, then proactively sense to maximize benefits and/or minimize costs. However, properties of complex environments make active sensing more difficult, necessitating further research and evaluation before deploying such solutions. In this paper, we describe MineralMiner, a novel simulation environment that extends previous environments to provide eight common complex environment properties in order to enable the effective study of active sensing. These properties can be fine-tuned using several simulation parameters in order to properly mimic environments likely to occur in real-world applications, allowing for insightful and successful pre-deployment testing, evaluation, and debugging of active sensing solutions, as well as on-going research into active sensing. Furthermore, we describe how several applications of sensing problems benefiting from active sensing can be abstracted and studied within MineralMiner, demonstrating the breadth and depth of its applicability to active sensing research.

Neural Networks for Survey Researchers

Neural networks are currently one of the most popular and fastest growing approaches to machine learning, driving advances in deep learning for difficult real-world applications ranging from image recognition to speech understanding in personal assistant agents to automatic language translation. Although not yet as commonly employed in survey research as other types of machine learning, neural networks offer natural extensions of well-known linear and logistic regression techniques in order to learn non-linear functions predicting or describing nearly any real-world process or problem (provided there are sufficient data and an appropriate set of parameters). Moreover, neural networks offer great potential towards more intelligent surveys in the future (e.g., adaptive design tailored to individual respondents’ characteristics and behavior, automated digital interviewers, analysis of rich multimedia data provided by respondents). Neural networks can learn for both regression and classification tasks without requiring assumptions about the underlying relationships between predictive variables and outcomes. In this article, we describe what neural networks are and how they learn (with tips for setting up a neural network), consider their strengths and weaknesses as a machine learning approach, and illustrate how they perform on a classification task predicting survey response from respondents’ (and nonrespondents’) prior known demographics.

An Introduction to Machine Learning Methods for Survey Researchers

Machine learning techniques comprise an array of computer-intensive methods that aim at discovering patterns in data using flexible, often nonparametric, methods for modeling and variable selection. These methods offer an expansion to the more traditional methods, such as OLS or logistic regression, which have been used by survey researchers and social scientists. Many of the machine learning methods do not require the distributional assumptions of the more traditional methods, and many do not require explicit model specification prior to estimation. Machine learning methods are beginning to be used for various aspects of survey research including responsive/adaptive designs, data processing and nonresponse adjustments and weighting. This special issue aims to familiarize survey researchers and social scientists with the basic concepts in machine learning and highlights five common methods. Specifically, articles in this issue will offer an accessible introduction to: LASSO models, support vector machines, neural networks, and classification and regression trees and random forests. In addition to a detailed description, each article will highlight how the respective method is being used in survey research along with an application of the method to a common example. The introductory article will provide an accessible introduction to some commonly used concepts and terms associated with machine learning modeling and evaluation. The introduction also provides a description of the data set that was used as the common application example for each of the five machine learning methods.

Understanding the Human Condition through Survey Informatics

Survey informatics leverages two separate but vital areas of research--survey methodology and computer science and engineering--to advance the state of the art in each and improve our understanding of the human experience.