Large language models (LLMs) are increasingly appearing in consumer-facing products. To prevent problematic use, the organizations behind these systems have put content moderation guardrails in place that prevent the models from generating content they consider harmful. However, most of these enforcement standards and processes are opaque. Although they play a major role in the user experience of these tools, automated content moderation tools have received relatively less attention than other aspects of the models. This study undertakes an algorithm audit of OpenAI’s ChatGPT with the goal of better understanding its content moderation guardrails and their potential biases. To evaluate performance on a broad cultural range of content, we generate a dataset of 100 popular United States television shows with one to three synopses for each episode in the first season of each show (3,309 total synopses). We probe GPT’s content moderation endpoint (ME) to identify violating content both in the synopses themselves, and in GPT’s own outputs when asked to generate a script based on each synopsis, also comparing with ME outputs on 81 real scripts from the same TV shows (269,578 total ME outputs). Our findings show that a large number of GPT-generated and real scripts flag as content violations (about 18% of GPT scripts and 69% of real ones). Using metadata, we find that TV maturity ratings, as well as certain genres (Animation, Crime, Fantasy, and others) are statistically significantly related to a script’s likelihood of flagging. We conclude by discussing the implications of LLM self-censorship and directions for future research on their moderation procedures.
Adaptive learning technologies offer personalized education by tailoring unique learning paths to individual students. Despite extensive study, widespread adoption remains limited due to the considerable costs associated with development and deployment. This paper presents the application of a design-based research framework, integrating Wizard of Oz techniques, intervention design, and decision trees to simulate an adaptive learning platform. This platform dynamically encourages collaborative efforts among students based on their social learning tendencies. Theories of social learning highlight the importance of social interactions in facilitating effective learning outcomes, particularly in remote learning environments. Our research methodology adopts an interdisciplinary approach, drawing insights from social learning theory to explore system concepts within a fluid framework. By employing this methodology, we seek to contribute to ongoing discussions aimed at enhancing technology-supported learning environments.
Adaptive learning systems aim to emulate how skilled educators seek to provide every student the best possible learning experience. We investigate how such systems might be enriched by activities and indicators of social learning - an aspect of learning that focuses on the influences of learners’ social context and interactions. In this paper we describe a pilot study aimed at exploring the potential for including social learning in an adaptive system. Our analysis of the social learning scale demonstrates its validity and usefulness for our ongoing work. Our qualitative analysis of students’ learning demonstrates how social learning vary among students. We discuss how the rating scale results and observations of social learning can be integrated within a student model needed to drive an adaptive system. More generally, our work illustrates how theories of learning can contribute to the design of adaptive learning systems.
We examine the major challenges AI experts encounter when deploying AI solutions. We conduct in-depth interviews with AI experts to elicit the frequent challenges they face. Our preliminary results highlight several factors that contribute to the challenging nature of the field, namely data scarcity, uncertainty, ill-structured problems, user behavior, and the application domain. This work indicates important areas that can be targeted during the curriculum design and development process.