Navigating the Cons of AI Use

Produce convincing but unreliable information

AI responses can contain false claims, invented citations, distorted summaries, and outdated information. Because they sound confident and polished, problems are hard to spot without checking other sources.

Students should evaluate AI output before relying on it. They can check important claims against course readings, library databases, original sources, and disciplinary expertise.

Replace productive struggle

Writing is one way students discover relationships, test ideas, identify gaps, work through uncertainty, and develop knowledge. When AI performs these tasks too early, students can receive a finished structure or answer before they have worked through the problem themselves.

Not every difficulty is educationally valuable. Some create unnecessary barriers; others are exactly what the assignment is designed to help students practice. Instructors can help students tell the two apart.

Flatten voice and linguistic difference

AI systems often produce standardized and formulaic prose. When students are encouraged to use AI to “fix,” “clean up,” or make their writing sound more “professional,” the resulting text may erase meaningful linguistic choices, cultural expression, disciplinary identity, or rhetorical experimentation.

Students can ask what a revision changes, whose ideas about “good writing” it reflects, and what is gained or lost when their language is rewritten.

Reproduce bias and omissions

AI systems learn patterns from uneven collections of data. Their responses can reproduce stereotypes, privilege dominant perspectives, leave out particular communities or forms of knowledge, and make assumptions about what counts as credible or relevant information.

These problems can appear in examples, descriptions of communities, recommendations, language judgments, and explanations of disciplinary or social issues. Students need practice judging how a response represents knowledge and whose perspectives it leaves out, in addition to whether it sounds plausible. Questions they can ask:

• Whose knowledge appears in this response?
• Whose knowledge is absent?
• What assumptions does the response make?
• How might another community or disciplinary perspective frame this differently?

Create privacy, consent, and intellectual-property risks

Using AI often involves sending information to an external system. Depending on the tool, account, settings, and type of data involved, prompts and uploaded files may be processed in ways students do not expect.

Students should not enter confidential information or upload someone else's writing without permission. This matters most when a task involves student drafts, peer writing, interview materials, research data, workplace documents, or other information that was not created for public distribution.

Increase surveillance and suspicion

Concerns about unauthorized AI use have also led to the adoption of AI-detection systems. These tools can produce false positives and false negatives, and research (opens in a new tab) has documented that some detectors disproportionately flag multilingual writers.

A detector score is not proof that a student used AI. When questions about authorship arise, instructors can look at the documented history of the student's work: drafts, notes, reflections, sources, and other evidence of the writing process.

Reinforce unequal access

Students do not have equal access to paid AI platforms, current devices, reliable internet, private spaces for completing coursework, or prior experience using AI tools. Requiring a particular commercial platform can create an additional barrier or pressure students to exchange personal data for access.

When a course requires AI use, check that students have an appropriate no-cost tool, and offer an equivalent alternative when possible.

Carry environmental and labor costs

AI systems depend on energy, water, hardware infrastructure, data labeling, content moderation, and other forms of human labor. These costs are largely invisible during an ordinary interaction with an AI tool and are not distributed equally across communities.

Instructors can invite students to reflect on when AI offers real educational value and when another approach would reach the same goal, and to decide whether to use these tools.