Sarah Goodwin

Sarah Goodwin

Auto-generated profile, based on publicly available information

Sarah Goodwin

Duolingo, Pittsburgh, PA, USA

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

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

Representative Publications

View All Publications

10 August 2025 in Language Testing
Verified
Journal Article
Open access
ARAndrew Runge
Andrew Runge
SGSarah Goodwin
Sarah Goodwin
YAYigal Attali
Yigal Attali
MPMya Poe
Mya Poe
PMPhoebe Mulcaire
Phoebe Mulcaire
GLGeoffrey T. LaFlair
Geoffrey T. LaFlair
A longstanding criticism of traditional high-stakes writing assessments is their use of static prompts in which test takers compose a single text in response to a prompt. These static prompts do not allow measurement of the writing process. This paper describes the development and validation of an innovative interactive writing task. After the test taker responds to an initial prompt, a large language model automatically analyzes their response to identify the prompt themes they discussed. The test taker then receives a customized follow-up prompt asking them to discuss one of the themes they did not address in their original response. Several experimental results support the task’s validity. Results showed that prompt phrasing is important to elicit responses most consistent with the task expectations and that test takers are able to produce responses that are relevant. Regarding issues of fairness in administration, we found that scores are not negatively affected by the type of follow-up task. We found that an accurate theme-detection model ensures test takers are not asked to write about something they addressed in their initial response, which could threaten test score validity claims. This design brings high-stakes writing tasks into closer alignment with writing models that frame writing as a process.

16 June 2022
Verified
Report
Closed access
YPYena Park
Yena Park
GLGeoffrey LaFlair
Geoffrey LaFlair
YAYigal Attali
Yigal Attali
ARAndrew Runge
Andrew Runge
SGSarah Goodwin
Sarah Goodwin
This paper introduces a new item type on the Duolingo English Test called Interactive Reading and grounds the item within the Duolingo English Test’s theoretical language assessment design framework and its assessment ecosystem. The innovative response format and automated item generation methods contribute to the specification of the construct of L2 reading, thereby strengthening the validity claims of the Duolingo English Test.

25 April 2023
Verified
Report
Closed access
GLGeoffrey LaFlair
Geoffrey LaFlair
ARAndrew Runge
Andrew Runge
YAYigal Attali
Yigal Attali
YPYena Park
Yena Park
JCJacqueline Church
Jacqueline Church
SGSarah Goodwin
Sarah Goodwin
This paper introduces a new integrated task type on the Duolingo English Test called Interactive Listening and grounds the task within the Duolingo English Test’s theoretical language assessment design framework and its assessment ecosystem. The task and automated item generation methods contribute to measurement of the constructs of L2 listening, reading, and writing, thereby strengthening the validity claims of the Duolingo English Test.

28 March 2023
Verified
Report
Closed access
SGSarah Goodwin
Sarah Goodwin
YAYigal Attali
Yigal Attali
GLGeoffrey LaFlair
Geoffrey LaFlair
YPYena Park
Yena Park
ARAndrew Runge
Andrew Runge
KYKevin Yancey
Kevin Yancey
Assessments, especially those used for high-stakes decision making, draw on evidence-based frameworks. Such frameworks inform every aspect of the testing process, from development to results reporting. The frameworks that language assessment professionals use draw on theory in language learning, assessment design, and measurement and psychometrics in order to provide underpinnings for the evaluation of language skills including speaking, writing, reading, and listening. This paper focuses on the construct, or underlying trait, of writing ability. The paper conceptualizes the writing construct for the Duolingo English Test, a digital-first assessment. “Digital-first” includes technology such as artificial intelligence (AI) and machine learning, with human expert involvement, throughout all item development, test scoring, and security processes. This work is situated in the Burstein et al. (2022) theoretical ecosystem for digital-first assessment, the first representation of its kind that incorporates design, validation/measurement, and security all situated directly in assessment practices that are digital first. The paper first provides background information about the Duolingo English Test and then defines the writing construct, including the purposes for writing. It also introduces principles underpinning the design of writing items and illustrates sample items that assess the writing construct.

27 July 2026 in Language Testing
Verified
Journal Article
Open access
YPYena Park
Yena Park
YAYigal Attali
Yigal Attali
XZXiaowan Zhang
Xiaowan Zhang
JCJacqueline Church
Jacqueline Church
KLKai-Ling Lo
Kai-Ling Lo
GLGeoffrey T. LaFlair
Geoffrey T. LaFlair
Despite task interactivity eliciting different aspects of language use, more interactive, adaptive speaking tasks have been challenging to implement in computer-delivered large-scale high-stakes contexts. Recent advances in large language models (LLMs) offer a potential means to address this tension through spoken dialogue systems (SDSs) but fall short of full adaptivity. To enhance adaptivity in an LLM-enhanced SDS-delivered interview task, this article proposes two solutions: (1) LLM-driven response evaluation using question-specific task completion rubrics for adaptive follow-up question selection and (2) item response theory (IRT) scoring accounting for prompt variability. We examined the extent to which these solutions functioned to simulate adaptivity, from 5,909 test takers’ responses to an adaptive speaking task with an avatar and a non-adaptive monologic task. Automated task completion evaluations were comparable to expert ratings, though follow-up question evaluation showed low agreement both among human raters and between raters and the system. IRT scoring yielded higher test–retest reliability, and the adaptive task elicited more reciprocal language use than the monologic task. Unlike earlier SDSs that prioritized consistency over adaptivity, the real-time meaning-level evaluation with IRT modeling strikes a balance between standardization and adaptivity. These findings support the viability of adaptive speaking tasks in computer-based standardized assessments.