fg-zettelkasten

A topic-anchored Zettelkasten built from the toread paper feed — 345 papers across 23 topics. The name follows the Niklas Luhmann archive model: each Topic is a Schlagwort (keyword) register that links back to every paper it covers. Use the graph, backlinks, and search to explore; the entry points below group the archive by theme.

To draft new writing from these notes with an AI agent, see Write with an AI agent.

Latest papers

Most recent additions to the feed.

Topics

Thematic registers — each lists every paper assigned to it.

  • Coordinated Inauthentic Behavior (57 papers) — Detection and analysis of coordinated link-sharing, inauthentic networks, and deceptive information operations on social platforms, including the CooRnet/CLSB methodology and the VERA-AI Alert monitoring system. This is the researcher’s longest-running and most active line of work, spanning election, health, and geopolitical case studies (Italy, Germany, Nigeria, India, Ukraine, Brazil).
  • Platform Governance & Data Access (72 papers) — Research and expert advisory work on platform policy, regulatory compliance, and researcher data access, including the Digital Services Act/ECAT expert role, analyses of Meta’s political content reduction policy, the CrowdTangle shutdown’s impact, and the state of social media research APIs post-DSA.
  • Generative AI & Disinformation (27 papers) — Emerging work on how generative AI is used in coordinated influence operations (e.g., AI-generated gambling promotion imagery) and how it reshapes disinformation tactics, paired with new teaching on generative AI and media aimed at training the next generation of researchers.
  • LLMs & Computational Content Analysis (30 papers) — Methodological research on integrating large language models into content analysis and annotation pipelines for political and social media discourse, including validation challenges, human-machine disagreement, and comparative embedding-model studies for clustering political text.
  • Italian Electoral Communication (MINE) (34 papers) — The MINE programme’s core empirical focus: mapping Italian news media and social media dynamics around national elections (2018, 2022) and related campaigns, including partisan attention, hyperpartisan networks, and computational measures of political parallelism.
  • Health Misinformation & Problematic Content (12 papers) — Studies of coordinated and organic circulation of problematic health information—COVID-19 fact-checking, vaccine discourse, and health content networks in India and Nigeria—examining online gatekeeping and post-truth counter-publics around medicine.
  • Political Polarization & Partisanship (79 papers) — Analysis of affective and structural polarization dynamics across social and traditional media, including partisan community engagement (e.g., pro-Bolsonaro networks in Brazil), cross-national comparisons in Latin America, and the interplay between social and legacy media in amplifying political extremes.
  • Information Disorder & Fake News (93 papers) — Foundational conceptual and taxonomic work on misleading information within hybrid media systems—defining ‘fake news,’ information disorder, and coordinated inauthentic behavior as theoretical categories that underpin the researcher’s broader empirical agenda.
  • Social TV & Second Screen (1 paper) — Earlier but recurring strand examining audience engagement with television via social media (Twitter second-screen behavior, social TV participation, streaming platform adoption), representing a historically significant but currently dormant research theme.
  • Sociocybernetics & Social Systems Theory (4 papers) — Theoretical work bridging sociocybernetics and internet/social systems theory, reflecting the researcher’s doctoral and early-career grounding in systems-theoretic approaches to digital society, still referenced in occasional ISA World Congress contributions.
  • Digital Research Methods & Ethics (36 papers) — Methodological contributions on collecting and analyzing social media data responsibly, including Facebook-ad-based survey recruitment, ethics of computational social science, and the limits/biases of platform-provided datasets for research.
  • Public and Sociotechnical Approaches to AI Safety Evaluation (5 papers) — Papers framing AI safety practices—red-teaming and evaluation—as sociotechnical activities requiring public involvement, labor consideration, and participatory science methods rather than purely technical procedures.
  • Online Radicalization and Extremism Across Platforms (20 papers) — Empirical and review studies of how extremist actors (far-right, jihadist) exploit platform affordances, algorithms, and multimodal content to spread and normalize radical ideologies online.
  • Retrospective Essays on Platformization and Social Media’s Trajectory (21 papers) — Reflective anniversary/position essays diagnosing the current state of social media and platform economies, tracing how platformization, political economy, and scam/extractive logics have reshaped online life over the past decade.
  • Public Attitudes and Labor Experiences with AI (13 papers) — Survey- and discourse-based studies of how publics and workers perceive, use, and are shaped by AI systems, including stakeholder discourse, national attitudes, and worker-reported productivity and displacement effects.
  • LLMs and Computational Methods for Research Workflows (10 papers) — Methodological papers deploying LLMs or fine-tuned models to augment scholarly tasks such as peer review, qualitative coding, and domain-specific classification, with attention to human-in-the-loop design and evaluation.
  • Message-Based Interventions for Climate Communication and Misinformation Resistance (11 papers) — Experimental and review studies testing how framed messages, chatbots, or inoculation techniques shape audience attitudes, resistance to misinformation, and engagement with climate or conspiracy content.
  • Survey Data Quality, Comprehension, and Recruitment (6 papers) — Papers examining the methodological foundations of online survey research, including whether respondents understand survey questions as intended, how representative and valid different opt-in sample sources are, and how novel platform-based recruitment strategies (e.g., Facebook ads) can be used and evaluated for reaching hard-to-survey populations.
  • Computational Approaches to Social Network Structure and Community Formation (20 papers) — Papers developing or applying computational/graph-based methods to detect, map, and explain emergent structural patterns—communities, practice clusters, and transitive ties—in large-scale online social networks, using techniques such as self-supervised graph embeddings and causal network analysis.
  • Cultural Imaginaries and Framings of Generative AI (11 papers) — Papers examining how generative AI is culturally imagined, discursively framed, and theorized as a medium, ranging from newspaper coverage and Silicon Valley community discourse to media-theoretical accounts of AI’s distinctive properties as a creative medium.
  • Recalibrating misinformation exposure and harms narratives (48 papers) — Papers that use empirical, large-scale, or naturalistic data to challenge alarmist public and scholarly assumptions about the prevalence, exposure patterns, and societal harms of online misinformation, showing that exposure is concentrated among few users, effects are limited, and everyday misinformation is broader and less exotic than commonly assumed.
  • Conspiracy Narratives Across Media Contexts (14 papers) — Studies examining how conspiracy theories and conspiratorial narratives are formed, represented, and circulated across different media environments—including journalism, alternative media networks, and image-centric social platforms like Instagram—focusing on the media logics and discursive strategies that shape conspiracy belief and narrative construction.
  • Performing Authenticity Under Platform Demands (7 papers) — Qualitative and mixed-methods studies examining how users negotiate authenticity, care, and legitimacy through performative practices shaped by platform affordances and demands — including micro-influencers’ branding of ethical care, Reddit users’ bot-like imitative posting, and TikTok influencers’ normalized incivility — arguing that these practices function as culturally negotiated, platform-specific survival or participatory strategies rather than fixed properties of sincerity or malice.

Structures

Curated narratives tracing a line of argument across the papers in a topic.

Write with an AI agent

zettel-paper is an Agent Skill — the open skill format used by Claude, OpenAI Codex, GitHub Copilot, Gemini CLI and other AI agents — that drafts papers, literature reviews and syntheses from fg-zettelkasten, the Luhmann way: it pulls a thread of linked notes that already forms an argument and turns it into prose, citing only papers in the archive, with their DOIs. It clones the public fabiogiglietto/fg-zettelkasten repository on every run, so it always works from the latest notes.

It works from the notes, not the papers: each paper’s summary, key claims, topics and links as published here. The full texts are not public, so by default the skill cannot read them. Treat a draft as a map of the literature, and check quotations, figures and fine-grained claims against the original paper before citing it.

Members with access to the kasten’s Google Drive folders can go further: in Claude, with the Google Drive connector signed in to an account the folders are shared with, ask for the full text of a named paper (“check the exact figure in …”) and the skill reads the PDF. Other agents have no such connector and stay on the notes.

Ask in plain language; the skill starts on intent. For example:

  • Draft a literature review on coordinated inauthentic behavior from the kasten.
  • What non-obvious paper could I write from these notes?
  • Turn the platform-governance Structure into a framing piece.

Install

Download the bundle: zettel-paper-skill.zip

  • Claude.ai, Claude Desktop and Cowork (Pro, Max): upload the zip as a personal skill in your Claude settings, under Capabilities → Skills. Code execution must be turned on there too.
  • Team and Enterprise: an organization admin uploads the zip once under Organization settings → Skills, and every member gets it.
  • Claude Code: clone fabiogiglietto/fg-zettelkasten and open it; the skill lives in .claude/skills/ and is found automatically. To have it in every project, unzip the bundle into ~/.claude/skills/.
  • OpenAI Codex CLI: unzip the bundle into ~/.codex/skills/.
  • Other agents: any tool that supports Agent Skills and gives the agent a shell with internet access (the skill clones the repository and runs a Python script). Unzip the bundle into that tool’s skills folder.

The skill was written and tested with Claude; other models can run it, but check their drafts with the same care.

Graph View