There is a compelling need to develop tools and strategies for people with intellectual and developmental disabilities (I/DD) in order to facilitate independence, self sufficiency, and address poor employment outcomes in adulthood. Through use of augmented reality (AR) and machine learning methods, we create an intelligent, contextually aware instructional system for persons with I/DD. We present results that demonstrate our system can be used independently by students with I/DD to quickly and easily acquire the skills required for performance of three relevant vocational tasks.

University lectures often offer online discussion forums for students to discuss and solve issues with other students and instructors. Correlating the participation of a student in a discussion forum to his performance in the course is subject of current research. Therefore, to qualify the different parts a student plays in a discussion, be it asking or answering a question, is sought in this paper. In current analysis of online discussion forums, such parts are annotated by hand. Thereby, identifying corresponding roles manually is a costly task, which requires the work of more than one person to annotate and approve the chosen roles. The desired step to a better understanding of student online discussion forums is the automated annotation of student roles. A student's role is determined by classifying the student's message into different speech act categories. This paper introduces a supervised speech act classification method for messages in German discussion forums that aims at solving the problem of manually detecting speech acts in online discussion for further discourse analysis. A comparative evaluation shows the significant improvements of the new classifier and its appropriateness for the German language.

Learners’ skills decay during gaps in instruction, since they lack the structure and motivation to continue studying. To meet this challenge, the PAL3 system was designed to accompany a learner throughout their career and mentor them to build and maintain skills through: 1) the use of an embodied pedagogical agent (Pal), 2) a persistent learning record that drives a student model which estimates forgetting, 3) an adaptive recommendation engine linking to both intelligent tutors and traditional learning resources, and 4) game-like mechanisms to promote engagement (e.g., leaderboards, effort-based point rewards, unlocking customizations). The design process for PAL3 is discussed, from the perspective of insights and revisions based on a series of formative feedback and evaluation sessions.

A thesis statement or controlling idea is a key component of the Common Core State Standards of writing from grade 6 to grade 12. We developed a machine learning model to identify thesis statements in students’ essays in order to focus peer-reviewers on commenting on the presence and quality of an author’s thesis statement. Identifying thesis statements in essays can be considered as a classification task in which a classifier is trained to predict whether a sentence is a thesis statement or not based on the features extracted from the sentence. However, the number of sentences in the thesis class is usually much lower than those in the not thesis class. Our initial model could not deal adequately with the challenge of class imbalance; there were too few instances of thesis statements from which to learn. Our subsequent model employs synthetic over-sampling in order to address this challenge and improve performance.

We present in this paper a study on negation in dialogues. In particular, we analyze the peculiarities of negation in dialogues and propose a new method to detect intra-sentential and inter-sentential negation scope and focus in dialogue context. A key element of the solution is to use dialogue context in the form of previous utterances, which is often needed for proper interpretation of negation in dialogue compared to literary, non-dialogue texts. We have modeled the negation scope and focus detection tasks as a sequence labeling tasks and used Conditional Random Field models to label each token in an utterance as being within the scope/focus of negation or not. The proposed negation scope and focus detection method is evaluated on a newly created corpus (called the DeepTutor Negation corpus; DT-Neg). This dataset was created from actual tutorial dialogue interactions between high school students and a state-of-the-art intelligent tutoring system.

The quality of student reflective responses has been shown to positively correlate with student learning gains. However, providing feedback on reflection quality to students is typically expensive and delayed. In this work, we automatically predict the quality of student reflective responses using natural language processing. With the long-term goal of producing informative feedback for students, we derive a new set of predictive features from a human quality-coding rubric. An off-line intrinsic evaluation demonstrates the effectiveness of the proposed features in predicting reflection quality, particularly when training and testing on different lectures, topics, and courses. An extrinsic evaluation shows that both expert-coded quality ratings and quality predictions based on the new features positively correlate with student learning gain.

Argument mining systems for student essays need to be able to reliably identify argument components independently of particular essay topics. Thus in addition to features that model argumentation through topic-independent linguistic indicators such as discourse markers, features that can abstract over lexical signals of particular essay topics might also be helpful to improve performance. Prior argument mining studies have focused on persuasive essays and proposed a variety of largely lexicalized features. Our current study examines the utility of such features, proposes new features to abstract over the domain topics of essays, and conducts evaluations using both 10-fold cross validation as well as cross-topic validation. Experimental results show that our proposed features significantly improve argument mining performance in both types of cross-fold evaluation settings. Feature ablation studies further shed light on relative feature utility.

Training scenarios, games, and learning environments often use narrative to manipulate motivation, priming, decision context, or other aspects of effective training. Computational representations of scenario narrative are useful for computer planning and real-time tailoring of training content, but typically define how to display narrative in the scenario world. The training rationales and the impacts of narrative on trainees are not typically accessible to the computer. We describe a computational representation that lets instructors explicitly author the training goals and impacts in a narrative. The representation captures both causality in the simulated world and instructional intent. Furthermore the streamlined representation enables non-technical authoring of sophisticated interactions between instructional goals when a computer tailors training material to individual learners. The narrative representation has the potential to increase instructor acceptance, understanding, and control over computer tailoring, thereby making training more effective.

While many expect that the use of advanced learning technologies like intelligent tutoring systems (ITSs) will substitute for human teaching and thus reduce the influence of teachers on student outcomes, studies consistently show that outcomes vary substantially across teachers and schools (Pane et al. 2010; Pane et al. 2014; Ritter et al. 2007a; Koedinger et al. 1997; Koedinger and Sueker 2014). Despite these findings, there have been few efforts (e.g., Schofield 1995) to understand the mechanisms by which teacher practices influence student learning on such systems. We present analyses of Carnegie Learning’s Cognitive Tutor ITS data from a large school district in the southeastern United States, which present a variety of usage and implementation profiles that illuminate disparities in deployments in practical, day-to-day educational settings. We focus on differential effectiveness of teachers’ implementations and how implementations may drive learner efficiency in ITS usage, affecting long term learning outcomes. These results are consistent with previous studies of predictors and causes of learning outcomes for students using Cognitive Tutor. We provide recommendations for practitioners seeking to deploy intelligent learning technologies in real world settings.

Hector Levesque has a strong critical position regarding the place of the Turing Test in Artificial Intelligence. A key argument concerns the test’s use of, or even, reliance on deception for subjectively demonstrating intelligence, and counters with a test of his own based on Winograd Schemas that he suggests is more objective. We argue that the subjectivity of the test is a strength, and that evaluating the outcome of Levesque’s objective test introduces other problems.