James C. Lester

Prof. James C. Lester


Research Focus & Metrics

Academic Career Length

20Years (2007-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
14 December 2007 in User Modeling and User-Adapted Interaction
Verified
Journal Article
Closed access
PDF file
SMScott W. McQuiggan
Scott W. McQuiggan
BMBradford W. Mott
Bradford W. Mott
JLJames C. Lester
James C. Lester

6 October 2023 in Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment
Verified
Journal Article
Open access
PDF file
VKVikram Kumaran
Vikram Kumaran
JRJonathan Rowe
Jonathan Rowe
BMBradford Mott
Bradford Mott
JLJames C. Lester
James C. Lester
Creating engaging interactive story-based experiences dynamically responding to individual player choices poses significant challenges for narrative-centered games. Recent advances in pre-trained large language models (LLMs) have the potential to revolutionize procedural content generation for narrative-centered games. Historically, interactive narrative generation has specified pivotal events in the storyline, often utilizing planning-based approaches toward achieving narrative coherence and maintaining the story arc. However, manual authorship is typically used to create detail and variety in non-player character (NPC) interaction to specify and instantiate plot events. This paper proposes SCENECRAFT, a narrative scene generation framework that automates NPC interaction crucial to unfolding plot events. SCENECRAFT interprets natural language instructions about scene objectives, NPC traits, location, and narrative variations. It then employs large language models to generate game scenes aligned with authorial intent. It generates branching conversation paths that adapt to player choices while adhering to the author’s interaction goals. LLMs generate interaction scripts, semantically extract character emotions and gestures to align with the script, and convert dialogues into a game scripting language. The generated script can then be played utilizing an existing narrative-centered game framework. Through empirical evaluation using automated and human assessments, we demonstrate SCENECRAFT’s effectiveness in creating narrative experiences based on creativity, adaptability, and alignment with intended author instructions.

12 June 2019 in IEEE Transactions on Learning Technologies
Verified
Journal Article
Closed access
WMWookhee Min
Wookhee Min
MFMegan H. Frankosky
Megan H. Frankosky
BMBradford W. Mott
Bradford W. Mott
JRJonathan P. Rowe
Jonathan P. Rowe
ASAndy Smith
Andy Smith
PSPeter Andrew Miller Smith
Peter Andrew Miller Smith
A distinctive feature of game-based learning environments is their capacity for enabling stealth assessment. Stealth assessment analyzes a stream of fine-grained student interaction data from a game-based learning environment to dynamically draw inferences about students' competencies through evidence-centered design. In evidence-centered design, evidence models have been traditionally designed using statistical rules authored by domain experts that are encoded using Bayesian networks. This article presents DeepStealth, a deep learning-based stealth assessment framework, that yields significant reductions in the feature engineering labor that has previously been required to create stealth assessments. DeepStealth utilizes end-to-end trainable deep neural network-based evidence models. Using this framework, evidence models are devised using a set of predictive features captured from raw, low-level interaction data to infer evidence for competencies. We investigate two deep learning-based evidence models, long short-term memory networks (LSTMs) and n-gram encoded feedforward neural networks (FFNNs). We compare these models' predictive performance for inferring students' knowledge to linear-chain conditional random fields (CRFs) and nave Bayes models. We perform feature set-level analyses of game trace logs and external pre-learning measures, and we examine the models early prediction capacity. The framework is evaluated using data collected from 182 middle school students interacting with a game-based learning environment for middle grade computational thinking. Results indicate that LSTM-based stealth assessors outperform competitive baseline approaches with respect to predictive accuracy and early prediction capacity. We find that LSTMs, FFNNs, and CRFs all benefit from combined feature sets derived from both game trace logs and external pre-learning measures.

1 November 2013 in International Journal of Artificial Intelligence in Education
Verified
Journal Article
Open access
PDF file
JSJennifer L. Sabourin
Jennifer L. Sabourin
LSLucy R. Shores
Lucy R. Shores
BMBradford W. Mott
Bradford W. Mott
JLJames C. Lester
James C. Lester
Self-regulated learning behaviors such as goal setting and monitoring have been found to be crucial to students’ success in computer-based learning environments. Consequently, understanding students’ self-regulated learning behavior has been the subject of increasing attention. Unfortunately, monitoring these behaviors in real-time has proven challenging. This paper presents an initial investigation into self-regulated learning in a game-based learning environment. Evidence of goal setting and monitoring behaviors is examined through students’ text-based responses to update their ‘status’ in an in-game social network. Students are then classified into SRL-use categories. This article describes the methodology used to classify students and discusses analyses demonstrating the learning and gameplay behaviors across students in different SRL-use categories. Finally, machine learning models capable of predicting these classes early in students’ interaction are presented.

30 January 2018 in IEEE Transactions on Learning Technologies
Verified
Journal Article
Closed access
ASAndy Smith
Andy Smith
SLSamuel Leeman-Munk
Samuel Leeman-Munk
ASAngi Shelton
Angi Shelton
BMBradford W. Mott
Bradford W. Mott
EWEric Wiebe
Eric Wiebe
PSPeter Andrew Miller Smith
Peter Andrew Miller Smith
Science learning is inherently multimodal, with students utilizing both drawings and writings to explain observations of physical phenomena. As such assessments in science should accommodate the many ways students express their understanding, especially given evidence that understanding is distributed across both drawing and writing. In recent years advanced automated assessment techniques that evaluate expressive student artifacts have emerged. However, these techniques have largely operated individually, each considering only a single mode. We propose a framework for the multimodal automated assessment of students' writing and drawing to leverage the synergies inherent across modalities and create a more complete and accurate picture of a student's knowledge. We introduce a multimodal assessment framework as well as two computational techniques for automatically analyzing student writings and drawings: a convolutional neural network-based model for assessing student writing, and a topology-based model for assessing student drawing. Evaluations with elementary students' writings and drawings collected with a tablet-based digital science notebook demonstrate that 1) each of the framework's two modalities provide an independent and complementary measure of student science learning, and 2) the computational methods are capable of accurately assessing student work from both modalities and offer the potential for integration in technology-rich learning environments for real-time formative assessment.

21 January 2021 in Journal of Science Education and Technology
Verified
Journal Article
Closed access
PDF file
ARArif Rachmatullah
Arif Rachmatullah
FRFrieda Reichsman
Frieda Reichsman
TLTrudi Lord
Trudi Lord
CDChad Dorsey
Chad Dorsey
BMBradford Mott
Bradford Mott
EWEric Wiebe
Eric Wiebe