Rhetorical AI Literacy : Human vs AI Generated Content

Introduction

Introduction

Hello everyone and welcome! My name is Kimberly Becker, and I am a co-founder of Moxy. This article expands on the third webinar in a series focusing on AI detection methods. If you're interested in learning more about Moxy, you can follow us on our various social media platforms.

Recap of the Webinar Series

This is the third webinar in a three-part series dedicated to understanding AI detection. In the first session, we discussed the fundamental building blocks of Large Language Models (LLMs) and their predictive algorithms. The second session took a deeper dive into the specifics of AI detection methods. Today, we will focus on rhetorical AI literacy by comparing human-generated writing with AI-generated content and discussing the limitations of AI detectors.

Characteristics of Human vs AI Writing: A Linguistic Perspective

Human Writing vs AI Writing

Having a background in Applied Linguistics, I aim to provide a linguist's perspective on AI detection and writing. We initially study human writing because AI-generated text models are based on patterns of human writing.

The Importance of Depth, Flow, and Accuracy

1. Depth:

   • The substance and complexity of the content.
   • Example: "Many countries are increasing their use of renewable energy" (surface level) vs "Many countries, including Sweden and China, are leading the transition towards renewable energy, embracing sources like wind and solar power."

2. Flow:

   • The smooth connection between ideas.
   • Example: Without flow: "Renewable energy is important. It can replace fossil fuels." With flow: "The importance of renewable energy cannot be overstated. It not only helps to replace fossil fuels but also leads to a more sustainable future."

3. Accuracy:

   • Correctness in terms of grammar, spelling, and citation format.
   • Example: "Renewable energy is good." (low accuracy) vs "Renewable energy sources like wind and solar have significant environmental benefits."

Characteristics of AI Writing

A study examined whether Turnitin could detect if texts were AI-generated. The detection was often inaccurate. The study identified certain patterns in AI-generated text:

1. Breadth vs Depth:

   • AI tends to offer broad, generalized statements without depth.
   • Example: "Climate change affects many regions" vs "Anthropogenic climate change, driven by greenhouse gas emissions, is a critical concern as evidenced in studies by IPCC (2021)."

2. Stance (Boosters vs Hedges):

   • AI writing often lacks nuanced stance markers. For instance, it uses more boosters and fewer hedges.
   • Example: AI: "The evidence clearly supports..." vs Human: "There may be a correlation between..."

3. Source Integrity:

   • AI writing often fails to integrate sources well.
   • Example: AI: "Innovative technologies can enhance outcomes" vs Human: "Johnson (2020) notes that innovative technologies enhance educational outcomes."

4. Rhetorical Situation:

   • Effective writing should address the audience and context but AI fails in achieving this effectively.

Rhetoric in Writing

Rhetoric revolves around effective communication and involves logical, emotional, and ethical appeals. In academic writing, rhetoric plays a crucial role. The goal of AI detectors is to distinguish between human and machine-generated content; however, considering rhetoric could make detection more effective.

Comparative Analysis:

• The study highlighted the differences between the vivid, audience-oriented human writing and the mechanical, formulaic AI-generated text.

Impact on Non-Native Speakers:

• AI detectors often flag non-native speakers' texts due to their predictability and formulaic usage.

Identifying Good Writing:

• Instead of detecting AI-generated text, focus on identifying effective writing characterized by depth, flow, and accuracy.

Faculty Guidance: Alternatives to AI Detection

• Consider approaches like in-depth analysis of students' understanding of content, including the extent of their engagement with rhetorical principles.

Keywords

Keyword

• AI Detection
• Human Writing
• AI Writing
• Rhetorical Literacy
• Depth, Flow, and Accuracy
• Stance in Writing
• Source Integrity
• Academic Rhetoric
• Large Language Models
• Formulaic Language

FAQ

FAQ

1. Why focus on human writing first?

   • Human writing provides the model from which AI-generated text is patterned. Understanding human writing characteristics is essential to identify deviations in AI writing.

2. What are the primary characteristics of human writing?

   • Depth (substance and complexity), Flow (smooth connection between ideas), and Accuracy (correct grammar, spelling, and citation).

3. How should AI detection be approached?

   • Instead of merely relying on AI detectors, focus on identifying effective writing which encapsulates depth, flow, and accuracy.

4. Why are non-native speakers more likely to be flagged by AI detectors?

   • Due to the formulaic and predictable nature of their language use, which are traits that AI detectors often interpret as characteristics of AI-generated text.

5. What role does rhetoric play in writing and AI detection?

   • Rhetoric is crucial for effective communication. Addressing audience, context, and purpose is vital and often overlooked in AI-generated text, giving clues to its origin.

6. What are alternatives to AI detection for faculty?

   • Assessing the depth of students' understanding, rhetorical engagement, and analyzing their use of sources can be more effective than relying solely on AI detectors.