(Re)learning critical reading in the age of GenAI

A key aim of many university modules is to develop critical reading strategies. This typically means taking the starting assumption that established published authors don’t necessarily have a fully coherent and objective world view. And that there are tensions and inconsistencies in their writing. The critical reader makes inferences about authorial intent.

The fundamental difference with AI-generated texts is that there is no authorial intent. Questions such as “Why did the author choose this phrase?” don’t make sense when analysing an AI-produced text. Or at least, they don’t meaningfully inform us as to the subjective world view of the author. Students therefore need to be re-equipped with critical reading strategies that serve them in the age of GenAI.

Step one: produce a relevant AI-generated text to read critically

The first step is to produce an AI-generated text that challenges the students’ critical reading skills. This means creative prompting, tweaking and exploring until you have produced a plausible, even passable, AI-generated essay. This works best with a topic that has been somewhat “AI-proofed” to require higher-level critical thinking skills, using phrases such as “critically evaluate”, “reflect on” and “from your reading of the course materials”.  The next steps involve systematically and critically interrogating this AI-generated text.

• Teaching students how to spot scientific misinformation
• Want to increase student satisfaction? Activate higher-order thinking skills
• Metacognition in education: get students thinking about their thinking

Step two: evaluate AI word choice

When critically reading human-written text, word choice is crucial for meaning. Ask students to highlight three or four key words used in the AI-produced text, for example, “freedom”, “rights”, “power”, “justice”. Ask them to evaluate how these words are used in the text. This will encourage students to evaluate the (lack of) authorial intent in the AI-produced essay and highlight alternative perspectives and understandings of key academic concepts.

Step three: evaluate AI retrieval

AI-generated texts are (in)famous for including “ghost references” and poor fact-checking. Ask students to choose four references or facts in the text to check the accuracy of (including page numbers, edition, and so on). This will encourage students to evaluate GenAI’s retrieval skills.

Step four: evaluate AI summary

Summary is a core critical reading skill that is undermined by AI tools. Prompt students to request “one-click” AI summaries of their readings. Encourage students to practise this skill by asking them to write their own summaries of their AI-produced texts. Then, ask GenAI to summarise the same text. Here, students can also compare texts generated on different AI platforms. How similar are the summaries? Which are more effective? Were the same perspectives or arguments emphasised? Can students critically evaluate how GenAI summarises texts? 

Step five: evaluate AI inference and bias

AI makes inferences by recognising patterns and drawing conclusions on unseen information by imitating how people reason and respond to prompts. As with human inference, these are deeply embedded with bias. It is now well documented that AI-generated texts consistently reflect numerous levels of bias. AI inferences, for example, will tend to favour the loudest, most prevalent internet sources. Encourage students to explore the inferences made in the AI-generated text by critiquing the arguments made. For example, “Can you come up with a counter-argument to the arguments made in the AI text?” This way, students must not only reveal some explicit and implicit bias in the AI-produced texts, but hone their critical reading skills to reveal how all inference is subjective and therefore subject to bias. 

Step 6: edit, improve and apply learning

This final step asks students to improve their AI-generated texts and develop their own arguments. Can students apply what they have learned about their AI-generated text to improve it? What will GenAI continue to do well and poorly? How to effectively prompt is a key transferable skill. Recognising the inevitable weaknesses of AI-generated texts is another. Now students have begun to more systematically and critically read their AI-generated texts, they can begin to make informed predictions about how GenAI will perform in different contexts. More pertinently, they will have practiced some crucial critical reading strategies to develop their own arguments.

GenAI allows students to bypass most steps of the critical reading journey. Their task becomes one of tweaking, developing and building on AI-produced texts. Yes, there is still space for critical reading, but the starting point is fundamentally different. Our students need to develop critical reading strategies to interrogate AI-produced texts if we hope to continue to nurture this key skill in the age of GenAI.

Brendan Carey is lecturer in politics in the department of humanities and social sciences at the University of Exeter, Penryn Campus.

If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.