Common Concerns About Generative AI

---

Generative AI models can sometimes produce text that is grammatically correct and highly convincing—but factually false. These outputs are known as hallucinations. The model is not intentionally lying; it simply generates language based on patterns it has seen, not verified knowledge.

   Example:
  A user asks, “Who discovered the element zirconium?” and the AI responds, “Zirconium was discovered by Marie Curie in 1923”—which is entirely incorrect. (The real answer: Martin Heinrich Klaproth, 1789.)

Inaccurate information can undermine learning, erode trust in educational tools, and lead to misinformation being repeated by students. Hallucinations are especially dangerous when learners are unfamiliar with the topic and assume the answer is correct.

At AI for Altruism, we:

• Use curated prompts and context constraints to reduce hallucinations in our applications.
• Combine LLMs with retrieval-augmented generation (RAG) to ground answers in verified sources.
• Clearly label experimental AI content and encourage users to cross-check claims.
• Maintain a human-in-the-loop philosophy: AI supports educators, never replaces expert judgment.
  

• Encourage students to verify AI outputs with credible sources.
• Use hallucinations as teachable moments—ask students to “fact-check the bot.”
• Consider classroom activities where students debunk an AI-generated claim.

With tools like ChatGPT, students can quickly generate essays, solve math problems, or write code—sometimes bypassing the learning process. This raises legitimate concerns about academic integrity.

• Teachers may struggle to assess authentic understanding.
• Over-reliance on AI can prevent students from developing foundational skills.
• Detection tools are imperfect—making fair assessment even harder.
  

• We promote AI literacy: helping students understand what AI is for, not just what it can do.
• We advocate for assignment redesign that values process over product (e.g., oral reflection, version tracking, creative synthesis).
• Our tools are built with transparency options: e.g., “Show Your Prompt,” “Explain Your Edits.”
• We support educators in crafting clear AI use policies in syllabi and academic codes of conduct.
  

• Open a dialogue: ask students when and why they use AI.
• Focus on tasks that emphasize reasoning, interpretation, and creativity.
• Assign collaborative or multi-modal projects that can’t be easily completed with AI alone.

Many educators worry that increasing reliance on AI in classrooms may lead to the devaluation—or even displacement—of their roles. The fear isn’t just about automation, but about erasing the human connection that makes teaching meaningful and effective.

• Teaching is more than delivering information; it involves empathy, adaptability, and mentorship.
• Systems that aim to “teach” using AI can sideline the nuanced, emotional, and cultural aspects of learning.
• The introduction of AI tools can lead to top-down decisions that ignore educator input, increasing burnout and mistrust.
  

At AI for Altruism, we design all tools with a teacher-first mindset. That means:

• AI supports workflow efficiency (e.g., drafting rubrics, summarizing feedback), not instruction replacement.
• Educators guide how AI is used and where it fits—they are the architects of AI integration, not passive recipients.
• We offer toolkits that require educator interaction and critical review, rather than “set-and-forget” automation.
• We actively advocate for teacher inclusion in edtech policymaking.
  

• Use AI to reduce administrative burden—e.g., auto-draft parent emails or generate assignment ideas.
• Explore AI as a collaborative teaching assistant, not an autonomous tutor.
• Push back against “displacement narratives” and instead co-design your classroom’s use of AI.
• Keep human connection front and center: no AI can replace the mentorship and insight you offer daily.
  

A 9th-grade English teacher uses AI to:

• Generate 3 example thesis statements for a novel study.
• Offer students real-time feedback on clarity—not grades.
• Save 3 hours/week on comments by using a structured prompt + personalized revision framework.
• Result: The teacher reports more energy for 1:1 conferencing and richer class discussions.

Generative AI systems are trained on vast amounts of data from the internet—data that reflects societal biases, historical inequalities, and cultural blind spots. As a result, AI outputs can reflect or even amplify biased, unfair, or inaccurate information.

   Example:
  An AI writing assistant might default to using male pronouns for doctors and female pronouns for nurses, even in gender-neutral contexts.

• Biased outputs can perpetuate stereotypes, undermine inclusion efforts, or subtly shape learners' worldviews.
• Inaccurate or one-sided explanations of history, culture, or science can lead to misinformation or cultural erasure.
• Educators may find themselves needing to “undo” or over-correct flawed content produced by AI tools.
  

At AI for Altruism, we take bias and accuracy seriously:

• We fine-tune system prompts to avoid harmful defaults.
• We provide transparency about how generative outputs are created and include disclaimers when AI is used.
• Where possible, we integrate retrieval-based methods to ground information in verifiable sources.
• We partner with educators from diverse backgrounds to test content across cultural contexts.
  

• Encourage critical media literacy: Ask students “Who might be missing from this narrative?” or “What assumptions are here?”
• Use biased AI outputs as springboards for discussion: Why did the system say this? What’s a more inclusive alternative?
• When using AI for content generation, check against primary sources or curated databases before presenting to students.
• Stay informed about AI ethics through groups like Algorithmic Justice League or Partnership on AI.
  
  

A history teacher asked an AI tool to “summarize causes of the Civil War” and received an answer focused almost entirely on “states’ rights.”

   ✅ Instead of discarding the tool, the teacher used it as a prompt: “Why do you think the AI missed slavery? What sources might it have overlooked?”

   Outcome: A powerful, student-led inquiry into historiography and digital literacy.

Generative AI tools are trained to predict likely responses to inputs, favoring clear, confident, and easily digestible answers. But not every educational topic lends itself to simplification—many require nuance, ambiguity, and critical reasoning.

   Example:
  Asked “Is capitalism good or bad?”, an AI may give a generalized, surface-level answer, glossing over economic theory, context, and counterarguments.

• Students may mistakenly believe that a single AI-generated answer reflects the whole truth.
• Oversimplification can reinforce passive learning and discourage exploration of complexity or contradiction.
• In fields like ethics, literature, science, and civics, a flattening of debate weakens understanding.
  

At AI for Altruism:

• We tune prompts to emphasize complexity and include qualifiers (e.g., “Explain multiple perspectives…”).
• Our tools suggest follow-up questions to encourage deeper exploration.
• We explicitly communicate: AI is a starting point, not the final word.
  

• Use AI outputs as first drafts for classroom discussion, not final answers.
• Ask students to critique AI explanations—what’s missing, misleading, or oversimplified?
• Design assignments where students must expand, revise, or rebut AI-generated content.
• Emphasize intellectual curiosity: “Good questions are better than perfect answers.”
  

In a philosophy class, students prompt AI with: “What is justice?”

AI provides a simplified summary of Rawls and Plato.
  Students then build comparative essays critiquing the gaps in the summary.

   Result: Deeper engagement with original texts and improved rhetorical reasoning.

As AI systems are introduced into more areas of daily life, there’s concern that meaningful human interaction is being replaced with scripted automation. From auto-generated condolences to AI companions marketed as friends or therapists, many worry that we’re outsourcing empathy and connection—and that institutions are using AI to avoid real investment in care.

• Human relationships are central to learning, growth, and healing. AI cannot replicate the emotional depth of real connection.
• When AI substitutes for communication (e.g., auto-responses from counselors or AI-led group discussions), students may feel unseen or dismissed.
• Using AI to “streamline” care interactions can unintentionally reinforce disconnection, especially for those who already feel marginalized.
• There’s a risk of normalizing emotional outsourcing, making it harder for students to build interpersonal trust and resilience.
  

At AI for Altruism, we draw a clear line: AI should never replace relational presence. Our approach includes:

• Building tools that support, not simulate, human interaction, e.g., helping educators prep for sensitive conversations, not deliver them.
• Requiring intentional review and customization before any AI-generated communication is shared with students or families.
• Advocating for policies that prioritize person-to-person support in mental health, counseling, and mentorship settings.
• Creating space for educators and students to co-reflect on when not to use AI.
  

• Use AI to prep notes or ideas before a 1:1 meeting, not to replace the meeting.
• Design activities where students reflect on AI’s limitations in emotional understanding.
• Keep eye contact, empathy, and active listening at the core of classroom culture.
• Challenge institutional uses of AI that feel emotionally hollow or impersonal.
  

A school counselor uses AI to:

• Draft a weekly check-in framework tailored to student themes.
• Track student reflections over time for patterns, not diagnoses.
• Brainstorm supportive language before making difficult calls to families.

   Result: The counselor feels more prepared, not replaced, leaving more energy for meaningful, face-to-face connection.

AI systems—especially those integrated into cloud platforms—often rely on collecting and processing user data to deliver responses or improve performance. In educational contexts, this raises urgent questions about student privacy, data rights, and ethical use.

• Educators may not always know what data is being stored or shared by third-party tools.
• AI interactions (even casual ones) can unintentionally include sensitive personal or academic information.
• Schools and districts are responsible for compliance with FERPA, COPPA, and other data regulations.
  

• We never store or reuse any user data for model training without consent.
• All data processing for our educational tools occurs in secure environments with data minimization principles.
• We support anonymous use and no-login access modes whenever possible.
• Our partners receive clear data usage agreements aligned with student safety and legal requirements.
  

• Check tool privacy policies before classroom use—look for red flags like “data sharing” or “training use.”
• Avoid entering student names, grades, or identifiers into AI tools.
• Teach students digital hygiene: never use real names or sensitive details in AI prompts.
• Work with your school’s IT team to vet tools for compliance with local and federal laws.
  

A middle school STEM teacher wanted to use AI to help students brainstorm project ideas.

Instead of having students prompt the AI directly, the teacher used a single shared account with anonymized prompts.

   Result: Student creativity soared—while privacy remained protected.