promptly.
← Semua materi
MATERI SINGKAT · 3 MENIT

Understanding AI Hallucinations

Learn why AI sometimes generates incorrect information, how to recognize hallucinations, and how to verify AI-generated responses using reliable sources.

Prinsip utama

AI hallucination occurs when an AI system generates information that appears convincing but is factually incorrect, misleading, or unsupported by reliable evidence. Large language models generate responses by predicting patterns in language. Although they can produce useful and accurate information, they do not automatically verify every statement against reliable sources. Hallucinations may occur when a question is ambiguous, information is missing, the model relies on unreliable patterns, or the user requests highly specific information that the model cannot reliably establish. Common examples include: 1. Fabricated academic references or research papers. 2. Incorrect historical dates or statistics. 3. Invented quotations attributed to real people. 4. Incorrect explanations presented with high confidence. 5. Nonexistent website links or sources. An important AI literacy principle is that a confident answer is not necessarily a correct answer. Users should apply critical thinking, verify important claims, and recognize uncertainty before relying on AI-generated information. Responsible AI use involves treating AI as an assistant rather than an unquestionable authority.

Contoh praktis

Scenario: A university student uses an AI chatbot to find academic references for a research assignment. Student's prompt: "Give me three academic papers proving that AI always provides accurate information." The chatbot generates three convincing academic references, including author names, publication years, and journal titles. However, when the student searches Google Scholar, one of the papers cannot be found. Problem: The AI may have fabricated a reference to produce a plausible answer. The student risks including nonexistent research in an academic assignment. Better approach: 1. Search for each paper using Google Scholar or the publisher's website. 2. Verify the title, author names, publication year, and DOI. 3. Read the original publication to confirm that it supports the claim. 4. Ask the AI to acknowledge uncertainty when reliable evidence is unavailable. Improved prompt: "Find peer-reviewed studies discussing the accuracy and limitations of AI-generated information. Provide verifiable references and explain any uncertainty." Lesson: Never assume that an AI-generated citation is authentic simply because it looks academically credible.

Coba sekarang

Exercise: Detect and Verify AI Hallucinations Objective: Practice evaluating AI-generated information before using it in academic work. Instructions: 1. Open an AI chatbot of your choice. 2. Ask: "List three peer-reviewed studies about the impact of artificial intelligence on university education." 3. Examine the references provided by the AI. 4. Search for each study using Google Scholar or an academic publisher's website. 5. Verify the title, authors, publication year, and DOI when available. 6. Identify any missing, incorrect, or unverifiable information. 7. Rewrite your prompt to request reliable sources and clear acknowledgment of uncertainty. Reflection questions: - Did the AI provide references that actually exist? - Were any details incorrect or unsupported? - Did the AI express uncertainty? - How could you improve the prompt? Expected outcome: Students understand that AI-generated information must be independently verified, especially when used for academic research.

Mulai Quiz →