Bias, bias everywhere: A response to Li et al. and Zhai and Nehm

Krist, Christina; Kubsch, Marcus

doi:10.1002/tea.21913

Cited by 5 publications

(6 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the context of machine learning, bias is defined as "undesirable system behaviors or properties manifesting itself in how the system exhibits different error rates for different kinds of language" (Cheuk, 2021, p. 826). To better understand the inherent bias in an AI system, Krist and Kubsch (2023) offer a useful framework by examining its lifecycle over three phases: inception, development, and deployment. During inception of the AI, bias may arise from the perspectives and priorities of the AI company's stakeholders in defining the AI's intended functions.…”

Section: Recognizing Bias In Genaimentioning

confidence: 99%

“…(Tang, 2020b). As the response originates from GenAI, an understanding of the inception, development, and deployment bias (Krist & Kubsch, 2023) is important to prompt critical questions, such as: "What are the sources of information for the AI? ", "what limitations exist in the AI's training data", and "What other theories or perspectives are left out in the explanation?"…”

Section: Connections To Genaimentioning

confidence: 99%

“…To the uninitiated, GenAI might seem like a "black-boxed" oracle that provides definitive answers to their questions. Such a perception will only reinforce the development and deployment biases highlighted by Krist and Kubsch (2023). Consequently, it is essential for students to understand GenAI's dialogic process in formulating responses and learn how to scrutinize the origins of GenAI's information and its transparency, or lack thereof, in source identification.…”

Section: Connections To Genaimentioning

confidence: 99%

See 2 more Smart Citations

Informing research on generative artificial intelligence from a language and literacy perspective: A meta‐synthesis of studies in science education

Tang

2024

Science Education

View full text Add to dashboard Cite

Research in languages and literacies in science education (LLSE) has developed substantial theoretical and pedagogical insights into how students learn science through language, discourse, and multimodal representations. At the same time, language is central to the functioning of generative artificial intelligence (GenAI). On this common basis concerning the role of language, this paper explores how foundational ideas from LLSE studies can inform the use of GenAI in science education. A bibliometric analysis of 412 journal articles from Web of Science provided the initial step to identify major themes and relationships in the LLSE literature. The analysis revealed four clusters of research in LLSE: reading and writing scientific text, science discourse and interaction, multilingual science classroom, and multimodality and representations. Each cluster was further analyzed through close reading of selected articles to identify and connect key constructs to the potential use of GenAI. These constructs include the interactive‐constructive reading model, text genre, reading‐writing integration, dialogic interaction, critical questioning, argumentation, translanguaging, hybridity, thematic pattern, modal affordance, and transduction. From these ideas and connections, the paper recommends several pedagogical principles for science educators to guide the use of GenAI. It concludes that LLSE research offers valuable insights for researchers and teachers to investigate and design the use of GenAI in science education. In turn, the impending use of GenAI also calls for a rethinking of literacy that will shape future research in LLSE.

show abstract

Section: Recognizing Bias In Genaimentioning

confidence: 99%

Section: Connections To Genaimentioning

confidence: 99%

Section: Connections To Genaimentioning

confidence: 99%

See 1 more Smart Citation

Informing research on generative artificial intelligence from a language and literacy perspective: A meta‐synthesis of studies in science education

Tang

2024

Science Education

View full text Add to dashboard Cite

show abstract

“…These beliefs are dangerous. Positioning science and data science as endeavors devoid of human decisions masks racist and other biased practices that may underlie these enterprises; for example, the use of racially biased datasets in predictive policing systems [12], hidden bias in AI systems for science education assessment that perpetuates inequities in education [13], and discriminatory designs in computer programming that unwittingly contribute to greater social inequity [14]. Professional learning that codifies the myth of data objectivity upholds the status quo in ways that reproduce educational injustices.…”

Section: Introduction 1overviewmentioning

confidence: 99%

“…Centering Voices involves addressing power and the power imbalance. 13. Centering Voices demands resources and long-term commitment.…”

mentioning

confidence: 99%

Centering Educators’ Voices in the Development of Professional Learning for Data-Rich, Place-Based Science Instruction

Wong,

Elsayed,

Nilsen

et al. 2024

Education Sciences

View full text Add to dashboard Cite

This self-reflective case study describes our project team’s efforts to promote equity in science professional learning (PL) by centering the voices of educators in the PL design process and within the course itself. We believe that educators’ experiences, priorities, and expertise are essential to developing professional learning that meets the needs of teachers and their students. We have a particular interest in amplifying the voices of those in historically underrepresented communities. Toward that end, we engaged science educators who work with Indigenous students and recent immigrants as collaborators in developing PL to support data-rich, place-based Earth Science instruction. In this case study, we share and critique the practices and tools that we have employed to center educator voices, rather than those of the PL designers and researchers. Our strategies for developing more equitable science professional learning include the use of: (a) equity-focused research methods, such as asset-based needs-sensing questions and peer-to-peer interviews; (b) a humanistic stance toward data-rich science learning, which emphasizes the typically unnamed sociocultural inputs and outputs that permeate all aspects of data; (c) a participatory design process that centers educator voices; and (d) a model of professional learning that uses representations of educator and student experiences as objects for reflection.

show abstract

Students’ Holistic Reading of Socio-Scientific Texts on Climate Change in a ChatGPT Scenario

Cheung,

Pun,

2024

Res Sci Educ

View full text Add to dashboard Cite

ChatGPT becomes a prominent tool for students’ learning of science when students read its scientific texts. Students read to learn about climate change misinformation using ChatGPT, while they develop critical awareness of the content, linguistic features as well as nature of AI and science to comprehend these texts. In this exploratory study, we investigated students’ reading performance in comprehending two ChatGPT-generated socio-scientific texts, with one focusing on cognitive-epistemic aspects of climate science and another one focusing on social-institutional aspects of climate science. We theorized such reading of ChatGPT-generated outputs as encompassing the content-interpretation, genre-reasoning and epistemic-evaluation domains. Combining Rasch partial-credit model and qualitative analysis, we explored and investigated how a total of 117 junior secondary students (grades 8 to 9) read such texts. Moreover, we also examined how 55 students’ holistic reading of socio-scientific texts on climate change in a ChatGPT scenario changes after a reading-science intervention. Our findings indicate that the content-interpretation was the easiest while the epistemic-evaluation domains were the most difficult. Interestingly, after the reading-science intervention, many students developed their tentative view on nature of science when they evaluated ChatGPT’s claims; while a small increase in number of students discussed reliability and non-epistemic nature of AI when they evaluated ChatGPT’s claims in relation to climate change. The findings also drive a pedagogical model that improves students’ holistic reading of socio-scientific texts generated by ChatGPT.

show abstract

Bias, bias everywhere: A response to Li et al. and Zhai and Nehm

Cited by 5 publications

References 23 publications

Informing research on generative artificial intelligence from a language and literacy perspective: A meta‐synthesis of studies in science education

Informing research on generative artificial intelligence from a language and literacy perspective: A meta‐synthesis of studies in science education

Centering Educators’ Voices in the Development of Professional Learning for Data-Rich, Place-Based Science Instruction

Students’ Holistic Reading of Socio-Scientific Texts on Climate Change in a ChatGPT Scenario

Contact Info

Product

Resources

About