As a Sophomore in Chinese Literature, I Had AI Read 96 Posts and Chose My Thesis Topic That Night

As a Sophomore in Chinese Literature, I Had AI Read 96 Posts
and Chose My Thesis Topic That Night

My thesis last semester nearly failed.

I spent two weeks negotiating the topic with my advisor. I downloaded 37 papers but only read abstracts. I struggled for a week to write under 3,000 words for the draft. Plagiarism check came back at 42%. I revised until 4 AM, ending up with text I couldn't even read smoothly myself. Three days before the deadline, I hadn't finalized the introduction.

Then my roommate handed me WorkBuddy. Desperate, I tried it for three days. I scored 85 on the thesis, with my advisor commenting, "Clear structure, sufficient argumentation."

I realized: Writing a thesis isn't about lack of effort; it's about tools mismatching your effort. Others are using AI to build frameworks, draft, reduce plagiarism scores, and format, while you're typing character by character.

Today, I'm sharing my complete 3-day workflow. Including pitfalls and failures, honestly.

I. Don't Rush Literature Reviews; Have AI Capture "What People Are Discussing"

Tool intro: WorkBuddy understands natural Chinese. Specify what to scrape, how to categorize, and tags, and it pulls social media discussions into a readable data table. No coding needed; even pure humanities students can handle it.

My bottleneck: I didn't know what real readers discuss regarding Dream of the Red Chamber. CNKI is full of academic chatter, not living voices. Surveys are slow; interviews are hard to arrange. So I gave this prompt:

Search Xiaohongshu posts from the last week. Keywords: Dream of the Red Chamber, Lin Daiyu, Xue Baochai, Character Discussions. Scrape post content and comments. Categorize by: Character Discussion, Plot Controversies, Film Adaptations, Social Etiquette. Extract title, summary, comment count, likes, publish time, IP location. Tag discussions. Finally, count by category.

I went to get water. Returned to find 96 entries listed: 32 Character Discussions, 24 Plot Controversies, 24 Social Etiquette, 16 Adaptations. Each had title, summary, metrics, time, location. AI added tags: Shipping CPs, Workplace Tactics, Art of Speech, Adaptation Complaints, Reading Notes.

One sentence, three minutes, 96 entries. Copy this step: Swap keywords for your field. Sociology -> Social Issues; Education -> Learning Struggles; Journalism -> Hot Events. Rest stays same.

II. Data Overturned My Assumed Topics

Don't use data blindly. First, I wrote down my three prior assumptions, then tested them against data. All three were wrong.

I assumed Baodai Romance was hottest. Yes, many posts mention Daiyu, but high-save-rate posts repeatedly cited by office workers were about "Xue Baochai's interpersonal tactics." Workers treat her as a social etiquette textbook.

I assumed high-likes would go to textual analysis. Instead, top-liked posts were mashups: "Red Chamber + Modern Context." Daiyu's speech in workplace communication; Baochai's strategy in dorm relationships. Pure textual discussion was minority.

I assumed discussants were peers/students. IP locations showed 60% in tier-1/2 cities. Active users resembled young professionals and exam candidates. Thus, discussing "Modern Reception" requires clear sample demographics, not assuming college students.

So I completely changed my topic: From "Causes of Tragedy in Baodai Romance" to "Modern Dissemination of Red Chamber Characters' Discourse—Case Study of Workplace Interpretations of Daiyu and Baochai." Both are about Red Chamber, but the latter rests on 96 real data points, giving me confidence.

III. AI Helps You Go from 0 to 1, But You Must Walk from 1 to 100

With data, I prompted AI to proceed. Second instruction:

Based on these 96 Red Chamber discussion data points, analyze: 1. Most discussed themes/high-frequency tags; 2. Common angles per theme; 3. Counter-intuitive findings; 4. Propose 5 viable thesis directions with data justification.

Among AI's suggested directions, one surprised me: "Modern Applicability of Daiyu's Art of Speech." Reason: Posts tagged "Art of Speech" had higher engagement. I hadn't considered this angle.

But three things AI cannot do, nor should you expect it to.

First, judge which direction is worthwhile. It listed 5, but "Does the 'Social Etiquette' data support a full thesis?