Skip to main content Skip to navigation

Challenging the appearance of machine intelligence: Cognitive bias in LLMs and Best Practices for Adoption

Project Overview

The document explores the role of generative AI, specifically large language models (LLMs), in education, focusing on their applications and the implications of cognitive biases inherent in these technologies. It highlights how LLMs can enhance learning experiences, facilitate personalized education, and support educators in decision-making processes. However, it underscores the critical importance of recognizing and addressing the cognitive biases present in these models to mitigate risks associated with their deployment. The document advocates for a cautious and informed approach to implementing LLMs in educational settings, emphasizing the necessity for transparency, accountability, and effective risk management strategies. Overall, while acknowledging the potential benefits of generative AI in fostering educational innovation, it calls for a balanced perspective that prioritizes ethical considerations and best practices to ensure positive outcomes for learners and educators alike.

Key Applications

Use of LLMs as decision support tools

Context: Workplace adoption of LLMs by consulting psychologists and educators

Implementation: Adoption of LLMs to assist in decision-making, while retaining human oversight

Outcomes: Increased efficiency and support in decision-making tasks

Challenges: Bias in outputs, reliance on LLMs as final decision-makers could lead to poor outcomes

Implementation Barriers

Technical barrier

Cognitive biases embedded in LLMs leading to inaccurate or biased outputs.

Proposed Solutions: Implement training programs to address cognitive biases, ensure oversight and review mechanisms.

Ethical barrier

Lack of transparency regarding the training data and potential biases in LLMs.

Proposed Solutions: Promote data literacy education and demand transparency from LLM developers.

Accountability barrier

Users are often held accountable for the outputs generated by LLMs, regardless of their accuracy.

Proposed Solutions: Establish clear guidelines and regulations for LLM usage in educational and professional settings.

Project Team

Alaina N. Talboy

Researcher

Elizabeth Fuller

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Alaina N. Talboy, Elizabeth Fuller

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies