About

Andrew S. Gordon is a Research Associate Professor of Computer Science and Director of Interactive Narrative Research at the Institute for Creative Technologies at the University of Southern California. His research focuses on advancing technologies for analyzing and generating narrative interpretations of time-series data, formalizing large-scale commonsense knowledge, and reasoning with these formalizations using logical abduction.

Research topics

• Computer Science
• World Wide Web
• Multimedia
• Art
• Human–computer interaction
• Artificial Intelligence
• Literature
• Programming language
• Linguistics
• Internet privacy

Selected publications

• Short, Animated Storytelling Video to Reduce Addiction Stigma in 13,500 Participants Across Multiple Countries Through an Online Approach: Protocol for a Randomized Controlled Trial (Preprint)

  2025-03-03

  preprint

  <sec> <title>BACKGROUND</title> Stigma toward people with addiction is a well-documented phenomenon that dramatically impacts help-seeking, treatment, and recovery. Interventions aimed at reducing stigma toward those with addiction must overcome the frequent mischaracterization of addiction as a failure of judgment rather than a chronic, treatable illness. Previous research has demonstrated that social contact with people recovering from addiction can promote empathy and reduce stigma, but social contact is difficult to scale. Short, animated storytelling (SAS) is a novel health communication approach that scales easily because it can leapfrog barriers associated with language, culture, literacy, and education levels. </sec> <sec> <title>OBJECTIVE</title> This study will investigate the effect of a cross-culturally accessible SAS video intervention aimed at reducing stigma and increasing empathy toward people with addiction. We also seek to gain insight into the mechanisms of action of this SAS intervention by measuring the contribution of sound design to their effect. </sec> <sec> <title>METHODS</title> We will conduct a randomized controlled trial with 13,500 adult participants from the United States, the United Kingdom, and South Africa, recruited online via Prolific Academic and randomized into 3 arms, per country. The 2 intervention arms will receive a wordless, social contact–based SAS video, one arm with a soundtrack and one without. The third arm will receive an educational video about addiction. Validated questionnaires will be used to assess our primary outcome, addiction stigma, and secondary outcomes, optimism, warmth toward the subject, and hopefulness, at baseline, immediately post exposure, and 2 weeks later. Ethics clearance was obtained on August 15, 2024, from the Stanford University institutional review board (protocol 76457). </sec> <sec> <title>RESULTS</title> This trial was funded in January 2025 by the Heidelberg Institute of Global Health, the Faculty of Medicine at Heidelberg University, in Germany. As of March 2025, no data have been collected. The estimated start date for this trial is May 15, 2025. We expect to complete data collection by July 1, 2025, and expect results to be published in the spring of 2026. </sec> <sec> <title>CONCLUSIONS</title> Here, we present the protocol for an online, multicountry, randomized controlled trial. This trial is designed to measure the effect of an innovative approach to global health communication (wordless, short, and animated storytelling) on addiction stigma in 3 global regions. These findings will inform the design of future scalable, digital health storytelling interventions for global audiences while exploring the capacity of SAS to shift public health attitudes and perceptions. Furthermore, if effective, the intervention described here could be disseminated broadly via social media and other online platforms. </sec> <sec> <title>CLINICALTRIAL</title> ClinicalTrials.gov NCT06705205; https://clinicaltrials.gov/study/NCT06705205 </sec> <sec> <title>INTERNATIONAL REGISTERED REPORT</title> PRR1-10.2196/73382 </sec>

  PublisherDOI

• Short, Animated Storytelling Video to Reduce Addiction Stigma in 13,500 Participants Across Multiple Countries Through an Online Approach: Protocol for a Randomized Controlled Trial

  JMIR Research Protocols · 2025-05-05 · 3 citations

  articleOpen access

  BACKGROUND: Stigma toward people with addiction is a well-documented phenomenon that dramatically impacts help-seeking, treatment, and recovery. Interventions aimed at reducing stigma toward those with addiction must overcome the frequent mischaracterization of addiction as a failure of judgment rather than a chronic, treatable illness. Previous research has demonstrated that social contact with people recovering from addiction can promote empathy and reduce stigma, but social contact is difficult to scale. Short, animated storytelling (SAS) is a novel health communication approach that scales easily because it can leapfrog barriers associated with language, culture, literacy, and education levels. OBJECTIVE: This study will investigate the effect of a cross-culturally accessible SAS video intervention aimed at reducing stigma and increasing empathy toward people with addiction. We also seek to gain insight into the mechanisms of action of this SAS intervention by measuring the contribution of sound design to their effect. METHODS: We will conduct a randomized controlled trial with 13,500 adult participants from the United States, the United Kingdom, and South Africa, recruited online via Prolific Academic and randomized into 3 arms, per country. The 2 intervention arms will receive a wordless, social contact-based SAS video, one arm with a soundtrack and one without. The third arm will receive an educational video about addiction. Validated questionnaires will be used to assess our primary outcome, addiction stigma, and secondary outcomes, optimism, warmth toward the subject, and hopefulness, at baseline, immediately post exposure, and 2 weeks later. Ethics clearance was obtained on August 15, 2024, from the Stanford University institutional review board (protocol 76457). RESULTS: This trial was funded in January 2025 by the Heidelberg Institute of Global Health, the Faculty of Medicine at Heidelberg University, in Germany. As of March 2025, no data have been collected. The estimated start date for this trial is May 15, 2025. We expect to complete data collection by July 1, 2025, and expect results to be published in the spring of 2026. CONCLUSIONS: Here, we present the protocol for an online, multicountry, randomized controlled trial. This trial is designed to measure the effect of an innovative approach to global health communication (wordless, short, and animated storytelling) on addiction stigma in 3 global regions. These findings will inform the design of future scalable, digital health storytelling interventions for global audiences while exploring the capacity of SAS to shift public health attitudes and perceptions. Furthermore, if effective, the intervention described here could be disseminated broadly via social media and other online platforms. TRIAL REGISTRATION: ClinicalTrials.gov NCT06705205; https://clinicaltrials.gov/study/NCT06705205. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): PRR1-10.2196/73382.

  PublisherDOI

• A short, animated storytelling video to reduce addiction stigma: A pilot randomized controlled trial

  Addictive Behaviors Reports · 2025-06-17 · 1 citations

  articleOpen access

  • Fully online pilot RCT demonstrates feasibility for large-scale stigma reduction study. • Single 2.5 min animated video significantly reduced addiction stigma post-view. • High retention (88%) and minimal data issues indicate digital trial feasibility. • Soundtrack and no-sound video arms showed similar immediate stigma reductions. • SAS videos can be scalable for social media to reduce public addiction stigma. Public stigma towards people with addiction negatively impacts help-seeking, treatment and recovery. This pilot study tested the feasibility of conducting a large-scale, online trial to measure the effect of a short, animated storytelling (SAS) stigma reduction video, with and without soundtrack, on addiction stigma, optimism, warmth towards people with addiction, and hopefulness at two timepoints (immediately post-exposure and 14 days later). We used a parallel group, three-arm randomized controlled trial (RCT). We conducted this fully online study on the Prolific Academic research platform (participant recruitment) and the Qualtrics survey platform (data collection). We recruited 631 English-speaking adult participants, aged 18–49, residing in the United States. Intervention group A received the SAS video intervention. Intervention group B group received the SAS video intervention without sound. The control group received written information about global addiction prevalence, estimated to be time-equivalent with the video interventions. We measured participant retention rate at the two-week follow-up to determine the feasibility of conducting the definitive trial. Our co-primary outcomes were addiction stigma, optimism, warmth towards people with addiction and hopefulness, measured using an abbreviated 18-item version of the Attribution Questionnaire (AQ-18), the Brief García’s Interactive Optimism Scale (BIOS-G), a stigma thermometer and a visual analogue scale (VAS). We used repeated-measures ANOVA to assess group-by-time interactions and compared changes from baseline to post-intervention. Participants completed follow-up surveys 14 days post-intervention. The retention rate from baseline to follow-up was 88.0 %. Exposure to both the video with sound and without sound resulted in significant positive changes compared to the control group, for pity [F (4,1046) = 3.26, η 2 = 0.012, p = 0.011], willingness to help [F (4,1046) = 8.48, η 2 = 0.031, p < 0.001], dangerousness [F (4,1046) = 2.95, η 2 = 0.011, p = 0.019], avoidance [F (4,1046) = 4.25, η 2 = 0.016, p = 0.002], as well as optimism [F (2,595) = 7.7, η 2 = 0.014, p < 0.001], warmth toward people with addiction [F (2,594) = 6.5, η 2 = 0.014, p = 0.002], and hopefulness [F (2,594) = 5.4, η 2 = 0.013, p = 0.005]. No effects were observed for fear or blame stigma sub-scales. These effects were no longer visible at follow-up in this pilot sample. No significant differences were observed between the video with sound and the video without sound. This pilot study demonstrates the feasibility of proceeding with our registered, largescale, multi-country, online RCT. The significant effect observed in a relatively small pilot population, after a single exposure to this 2.5 min SAS intervention aimed at reducing addiction stigma, was unanticipated and is worthy of highlighting. A larger sample size will adequately power the full trial to detect both immediate effects and their potential durability over time, in various global settings.

  PublisherDOI

• Embryology of a Language Model

  ArXiv.org · 2025-08-01

  preprintOpen access

  Understanding how language models develop their internal computational structure is a central problem in the science of deep learning. While susceptibilities, drawn from statistical physics, offer a promising analytical tool, their full potential for visualizing network organization remains untapped. In this work, we introduce an embryological approach, applying UMAP to the susceptibility matrix to visualize the model's structural development over training. Our visualizations reveal the emergence of a clear ``body plan,'' charting the formation of known features like the induction circuit and discovering previously unknown structures, such as a ``spacing fin'' dedicated to counting space tokens. This work demonstrates that susceptibility analysis can move beyond validation to uncover novel mechanisms, providing a powerful, holistic lens for studying the developmental principles of complex neural networks.

  PublisherOA PDFDOI

• From Test-Taking to Test-Making: Examining LLM Authoring of Commonsense Assessment Items

  2024-01-01

  articleOpen accessSenior author

  LLMs can now perform a variety of complex writing tasks.They also excel in answering questions pertaining to natural language inference and commonsense reasoning.Composing these questions is itself a skilled writing task, so in this paper we consider LLMs as authors of commonsense assessment items.We prompt LLMs to generate items in the style of a prominent benchmark for commonsense reasoning, the Choice of Plausible Alternatives (COPA).We examine the outcome according to analyses facilitated by the LLMs and human annotation.We find that LLMs that succeed in answering the original COPA benchmark are also more successful in authoring their own items.Premise: The girl received a trophy.What was the cause of this?Alternative 1: She won a spelling bee.Alternative 2: She made a new friend.Premise: I tipped the bottle

  PublisherOA PDFDOI

• From Test-Taking to Test-Making: Examining LLM Authoring of Commonsense Assessment Items

  arXiv (Cornell University) · 2024-10-18

  preprintOpen accessSenior author

  LLMs can now perform a variety of complex writing tasks. They also excel in answering questions pertaining to natural language inference and commonsense reasoning. Composing these questions is itself a skilled writing task, so in this paper we consider LLMs as authors of commonsense assessment items. We prompt LLMs to generate items in the style of a prominent benchmark for commonsense reasoning, the Choice of Plausible Alternatives (COPA). We examine the outcome according to analyses facilitated by the LLMs and human annotation. We find that LLMs that succeed in answering the original COPA benchmark are also more successful in authoring their own items.

  PublisherOA PDFDOI

• Combining the Predictions of Out-of-Domain Classifiers Using Etcetera Abduction

  2024-03-13

  article1st authorCorresponding

  Research in machine learning on Domain Adaptation has led to numerous methods for re-purposing highperformance pre-trained models for novel tasks, e.g., via finetuning a model with out-of-domain training data. When model weights are unavailable or otherwise fixed, there are fewer options available for exploiting its predictive power. In this paper we investigate whether the predictions of ensembles of fixed, pre-trained, out-of-domain image classification models can be used to improve the performance of an in-domain classifier, or replace it outright with comparable performance. Our approach involves computing the conditional probabilities from the confusion matrixes of out-of-domain predictions for in-domain training samples, then combining this information with prior probabilities and classification confidence using probability-ordered logical abduction, Etcetera Abduction, to select the most likely label for an in-domain test sample. We evaluate this approach using four image classification models in highly disparate domains. Results indicate that this method may be well-suited to applications where insufficient training data is available to train an accurate model on a novel task.

  PublisherDOI

• Playing Story Creation Games with Large Language Models: Experiments with GPT-3.5

  Lecture notes in computer science · 2023 · 9 citations

  Senior authorCorresponding
  • Computer Science
  • Computer Science
  • Artificial Intelligence
  PublisherDOI

• Searching for the Most Probable Combination of Class Labels Using Etcetera Abduction

  2023-03-22 · 2 citations

  article1st authorCorresponding

  Many machine perception tasks require a trained model to assign class labels to multiple entities in the same context, e.g., labeling multiple objects in a single photograph. In these tasks, different combinations of labels may be more likely than others, e.g., when co-occurrence biases are considered, such that the most-confident label assigned to an individual object is not always the best choice. In this paper, we propose a new method for combining evidence from multiple class probability distributions to identify the most probable combination of labels in multi-entity contexts. Our method encodes discrete class probability distributions as literals in first-order logic, and uses probability-ranked logical abduction to identify the most likely label combination, incorporating the prior and conditional probabilities of each label. We evaluate our method on two computer vision benchmarks, first for labeling common objects in photographs of everyday contexts, and second for labeling actions of athletes in sports videos. Results indicate significant gains in classifier accuracy over systems that merely select the model's most confident class label.

  PublisherDOI

• Demonstrating the Rapid Integration &amp; Development Environment (RIDE): Embodied Conversational Agent (ECA) and Multiagent Capabilities

  2022-05-09

  article

  We demonstrate the Rapid Integration &amp; Development Environment (RIDE), a research and development platform that enables rapid prototyping in support of multiagents and embodied conversational agents. RIDE is based on commodity game engines and includes a flexible architecture, system interoperability, and native support for artificial intelligence and machine learning frameworks.

  PublisherDOI

Frequent coauthors

• Jerry R. Hobbs

  180 shared

• Reid Swanson

  University of Southern California

  56 shared

• Melissa Roemmele

  55 shared

• Kenji Sagae

  26 shared

• Christopher Wienberg

  University of Southern California

  19 shared

• Andrew Feng

  Creative Technologies (United States)

  12 shared

• Naoya Inoue

  11 shared

• Olivia Connolly

  University of Southern California

  9 shared

Education

• Ph.D.

  Northwestern University

  1999