Rui (Ash) Feng
ash.rui.feng@gmail.com ·ashfeng.com ·Nagoya, Japan
Research Interests
Designing and testing media-system interventions against polarization. Current work targets perceived polarization, the misreading of what the other side actually thinks, deriving intervention modules from documented mechanisms of out-group misperception and implementing them as feed and interface conditions in a platform-independent experimental simulator. The next step is controlled online tests of their causal effects; conversational AI is a parallel arena for the same logic.
Education
M.A. in Sociology (Mass Communication), highest grade in all graded coursework
Thesis (in Japanese): The Role of the People's Daily WeChat Account in Political Propaganda — With a Focus on Patriotic Propaganda
B.A. in English Language and Literature
Exchange Student
Research Experience
Independent research project, sole researcher and developer — lichtung.ashfeng.com · github.com/rfb4c/Lichtung · video walkthrough
- — Designed and built a platform-independent social feed simulator for research on perceived polarization, demonstrating the interventions as interactive feed- and interface-level modules and running without access to commercial-platform APIs, hence without dependence on platform cooperation.
- — Derived three intervention modules from documented mechanisms of out-group misperception and implemented each: reranking that surfaces counter-stereotypical out-party content, a distribution-overlay UI that surfaces the moderate majority obscured by extreme voices, drawing on authoritative poll distributions, and an identity layer that displays shared non-political attributes.
- — Extending it for online experiments, adding randomized condition assignment, behavioral telemetry (exposure sequence, dwell time, scroll depth, interactions), and pre/post survey linkage.
Prior independent research project, sole developer — ashfeng.com/research
- — Built and deployed a news aggregator that collected coverage of the same event from outlets across the political spectrum, premised on the assumption that breaking filter bubbles would reduce polarization.
- — Shut it down after a systematic literature review found the filter-bubble account weakly evidenced (Bruns, 2019; Arguedas et al., 2022) and cross-cutting exposure capable of backfiring (Bail et al., 2018), moving the research target from what users see to how they interpret it; led directly to Lichtung.
Developed from M.A. thesis (Kansai University, 2022).
- — Hand-coded every headline the People's Daily WeChat account published January–April 2020 (N = 2,318; 761 coded as patriotic propaganda) under an original two-level scheme separating narrative direction (inward vs. outward) from construction mechanism.
- — Construction stayed inward-facing across all four crisis periods (94.1% at peak), and in the outward remainder positive self-positioning and antagonistic othering barely co-occurred (2.5% of 162), splitting the two halves of the standard in-group/out-group script.
- — Horizontal affective solidarity was the largest of five construction mechanisms (48.9% of the patriotic sample) and ran on a reusable heroization template: three formulaic title patterns and three narrative paths, including 28 posts in 16 days sharing one provincial medical-team roster format, with represented figures converging on healthcare workers.
Journalism
Journalist
Reported on government, science and technology, and general news.
- — Investigation into employment discrimination against people recovering from COVID-19. Prompted a policy response from the Ministry of Human Resources and Social Security.
- — National infrastructure feature series. Member of the feature team; the series later received the China News Award.
Intern Journalist
Covered cultural affairs in writing, photography, and documentary video.
- — Documentary on Chinese students in Japan during the COVID-19 pandemic. Independently reported, shot, and edited.
- — Photography desk and interactive VR dispatches. Profiles of documentary photographers and exhibition criticism; photo essays and VR reports on pandemic-era Shanghai and urban redevelopment.
Industry Experience
- — Retrieval architecture over system documentation and code, major energy utility (August 2026–Present). Proposal team member; running the client requirements sessions and drafting the technical approach: hybrid retrieval across the system's full artifact chain — requirements, design documents, source code — combining a vector store with a knowledge graph of the traceability relations between them.
- — Content-generation AI agent for a municipal government's international public relations division (April–June 2026). Developer, then recovery lead; team of 10. Architecture and implementation in Python and Strands Agents: prompt design and generation pipeline.
- — AI-driven development workflow for a major energy utility (August 2025–July 2026). Development lead, team of 5. Designed the AI workflow architecture and MCP integration for automated generation of Java error-handling code in a large-scale customer information system.
Full-stack agile development of a vehicle design support system for a major automotive manufacturer: detailed design through implementation, unit test, and system test in Java, TypeScript, and Oracle.
Awards and Honors
2025 · Internal MVP Award, ITXO GenAI Center, Accenture Japan (generative-AI engagement for a major energy utility)
2023 · China News Award (中国新闻奖) — the national journalism award of China (All-China Journalists Association); team award, national infrastructure feature series, The Cover
Skills
Certifications and Languages
2026 · Google Cloud Generative AI Leader
2025 · AWS Certified Cloud Practitioner
2019 · Japanese Language Proficiency Test, N1
Chinese (native); English (fluent); Japanese (fluent)