Hard vs. Easy Tasks with AI¶
The hard-vs-easy inversion describes the counterintuitive phenomenon where AI makes traditionally difficult tasks (writing code, literature synthesis) relatively easy while struggling with tasks humans find trivial (counting, precise formatting, basic arithmetic).
Context & Background¶
Traditional difficulty hierarchies in academia — where writing a literature review is harder than checking a citation — are inverted by AI capabilities. This creates a disorienting experience for new AI users who expect AI to handle simple tasks reliably before attempting complex ones.
Examples of the inversion:
- "Hard" tasks AI handles well: Drafting papers, writing complex code, synthesizing multiple sources, translating between programming languages
- "Easy" tasks AI struggles with: Precise counting, exact formatting, maintaining consistency across long documents, basic arithmetic without tools
Practical Implications¶
- Don't start with "easy" tasks to test AI: You might wrongly conclude it's useless
- Assign cognitively complex tasks: AI often excels at synthesis, translation, and generation
- Verify simple outputs extra carefully: Paradoxically, AI's simple outputs need more checking
- Use tools for precision: Give AI access to calculators and code execution for exact computation
Key Sources¶
- Voice Dictation
- Using LLMs with Cursor: Modern AI for Economics Research
- My Claude Code Setup - Academic Workflow Template
- Generative AI for Economic Research: Use Cases and Implications for Economists
- A Guide to Which AI to Use in the Agentic Era