Prompt Engineering as the New Core Literacy: Findings from a 243-User Survey on LLM Productivity

Forum topic · QianXun · 2025-11-20

Summary

A long-form Chinese forum post reviews Rizal Khoirul Anam's 2025 study (arXiv:2507.18638) on prompt engineering and large language model (LLM) productivity, based on a six-week descriptive quantitative survey of 243 global users conducted via Google Forms. Key findings: respondents were highly educated (65.5% held a bachelor's degree or above) and multidisciplinary, with psychology, engineering, and medicine leading; over 66% used AI tools at least twice weekly, primarily for academic/professional writing (165), summarizing (142), and coding (101). The most popular techniques were role prompting (105 users), chain-of-thought prompting (97), instruction prompting (94), and zero-/few-shot prompting (93/89). Over 55% of users frequently revised their prompts, reflecting metacognitive, human-in-the-loop interaction. Notably, 83.7% agreed that clearer, more specific prompts yield better results, and 75.7% reported faster task completion; education level correlated positively with prompt-strategy diversity, raising concerns about amplified digital inequality. The article argues prompt engineering is becoming a core digital literacy akin to typing or web search, urging educational and corporate integration of AI-collaboration skills.

Introduction

This Chinese forum post presents an in-depth commentary on *Prompt Engineering and the Effectiveness of Large Language Models in Enhancing Human Productivity* by Rizal Khoirul Anam (Nanjing University of Information Science and Technology), arXiv preprint arXiv:2507.18638v2 (2025). Using the metaphor of a sleeping giant awakened by well-crafted words, the article frames prompt engineering as the "alchemy" that unlocks large language model (LLM) productivity.

Study Design

Method: Descriptive quantitative research via a structured online questionnaire
Data collection: Google Forms, January 5 – February 10, 2025 (six weeks), distributed through WhatsApp, Telegram, LinkedIn, and Discord
Sample: 243 valid responses from a deliberately diverse, global population (China, Germany, Brazil, Vietnam, the US, Nigeria, Indonesia, Turkey, the UK, and more)

Who Is Using AI?

Education: 65.5% held a bachelor's degree or higher (34.2% bachelor's, 24.3% master's, 7% PhD) — an irony given AI's "democratization" narrative
Gender: nearly balanced (34.6% male, 32.1% female, 33.3% undisclosed/other)
Disciplines: psychology (31), engineering (28), and medicine (28) led; computer science (25) ranked only fourth — AI has spread far beyond programming circles

Usage Patterns

Over 66% (160 users) used AI at least twice weekly; 32 were daily heavy users
Top tasks: academic/professional writing (165), summarizing and paraphrasing (142), coding/code generation (101), creative content (89), data analysis and visualization (83), tutoring (62), translation (56)

Prompting Strategies in the Wild

Popularity ranking of self-adopted techniques:

1. Role prompting (105 users) — assigning the AI a persona, e.g., "you are an experienced marketing director" 2. Chain-of-thought prompting (97) — asking the model to reason step by step 3. Instruction prompting (94) — clear, specific, unambiguous directives 4. Zero-shot prompting (93) — direct task execution without examples 5. Few-shot prompting (89) — providing demonstrations for the model to imitate

The article notes these mirror the fundamentals of high-quality human communication: role-play, stepwise reasoning, precise instructions, and teaching by example.

Iteration as Metacognitive Skill

More than 55% of respondents (136) often or occasionally revised their prompts. The author describes this as "metacognitive interaction" — users act as active "conversation sculptors" in a human-in-the-loop loop, refining intent through successive rounds. Frequent prompt revision correlated with significantly higher satisfaction.

Education and Prompt Sophistication

Bachelor's holders used zero-shot (40), few-shot (30), instruction (27), role (12), and automatic prompt engineering (5) techniques
High-school graduates used far fewer: 18, 10, 11, 4, and 1 respectively

The post warns that prompt engineering may become a new form of "cultural capital," potentially amplifying existing educational inequality.

Productivity Impact

83.7% (203 users) agreed that clearer, more specific prompts yield better AI results
75.7% (184) confirmed AI helped them complete tasks faster
Mean ratings: 4.01/5 for prompt impact on output quality; 3.87/5 for AI-supported work efficiency — and these were strongly positively correlated

Conclusions

The article concludes that productivity gains in the AI era depend less on raw model scale than on human skill in crafting language. Prompt engineering is transitioning from a niche skill to a core literacy — comparable to typing (PC era) and web search (internet era). The author calls for AI-literacy curricula in secondary schools, corporate prompt-engineering workshops, prompt knowledge bases, and treating prompt quality as a managed process metric.

References

1. Anam, R. K. (2025). *Prompt Engineering and the Effectiveness of Large Language Models in Enhancing Human Productivity*. arXiv preprint arXiv:2507.18638v2. 2. Brown, T., et al. (2020). *Language Models are Few-Shot Learners*. NeurIPS 33, 1877–1901. 3. Wei, J., et al. (2022). *Chain-of-Thought Prompting Elicits Reasoning in Large Language Models*. NeurIPS 35, 24824–24837. 4. Liu, P., et al. (2023). *Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in NLP*. ACM Computing Surveys, 55(9), 1–35. 5. Bommasani, R., et al. (2021). *On the Opportunities and Risks of Foundation Models*. arXiv preprint arXiv:2108.07258.