Gabriele de Seta

Before the research question: computers, Chinese and music

An Italian born in 1986, de Seta belongs to the generation whose early Internet experience preceded today’s platform vocabulary.³ His own memories include learning HTML and DOS during summer courses with his grandfather, building a Warcraft III clan website and spending years on hardware forums and music-magazine discussion boards.⁴ Computing already involved making things for other people and inhabiting specialist communities. It was a practiced part of life before becoming a professional subject.

He studied Eastern Languages and Cultures at Sapienza in Rome from 2005 to 2009, then Chinese Studies at Leiden from 2009 to 2011.⁵ He later explained the original choice through the promise that China represented the future. By the end of the first degree, he wanted to move beyond its classical emphasis toward contemporary life.⁶ The language remained useful after the promise itself became questionable: it opened conversations in which the people concerned could complicate an outsider’s account.

His 2011 Leiden thesis, Mediation through Noise, investigated experimental music in China. Maghiel van Crevel and Jeroen Groenewegen supervised it; Groenewegen’s expertise in Chinese popular music was especially important. An interview appendix records encounters with musicians including Junky, Wu Quan, Yan Jun and Ronez, in Chinese and English, across Shanghai and Beijing from 2007 onward.⁷ Musical interest was acquiring scholarly procedures through a language program, specialist guidance and repeated contact.

The thesis followed a social practice rather than simply identifying a repertoire. Amateur learning, modified instruments, performances, recordings and online exchanges helped sustain the music.⁸ De Seta later identified the Shanghai research as the setting in which musicians’ use of local applications drew him toward digital media.⁹ He did not need to leave those relationships to find another object. The musicians had information to circulate, performances to arrange and friendships to maintain.

Reading Olia Lialina and Dragan Espenschied’s Digital Folklore in 2011 gave him a possible name for this ordinary creativity. Its attention to amateur websites resonated with his own Italian Internet world. But when he proposed Chinese digital folklore for doctoral study, sociology interlocutors rejected the category as academically unsuitable. He adopted vernacular creativity, drawing on Henry Jenkins and Jean Burgess.¹⁰ That accommodation would later become part of the history of his own research program.

Hong Kong: participating in two overlapping worlds

De Seta entered sociology doctoral study at Hong Kong Polytechnic University in 2012. He submitted Dajiangyou: Media Practices of Vernacular Creativity in Postdigital China in 2015 and received the PhD in 2016.^{11, 12, 13} Dajiangyou, literally buying soy sauce, can colloquially decline involvement in someone else’s affair. It suited an inquiry into everyday activity wary of assigning every Chinese Internet user a grand political purpose.¹⁴

He also entered Hong Kong’s experimental music world. On 10 October 2012, Shanghai contacts introduced him to Dennis Wong; de Seta brought a guitar and effects pedals to C.I.A. in an industrial building in Kwai Hing. Huang Lei later invited him to play, and he made flyers and helped publicize Strategic Sounds events. Over three years he performed in eight venues, seven in industrial buildings, repeatedly disrupted by rents and closures.¹⁵

A scene could persist while its rooms disappeared. His account of a no-venue underground arose from people continuing to perform, organize and communicate under those conditions.¹⁶ This was knowledge acquired as a musician with things to contribute, rather than through an external survey of available venues. It also kept music’s practical economy in view: relationships and enthusiasm did not remove the need for access to space.

Computing skills connected him to scholarly organization too. The 2014 Chinese Internet Research Conference at PolyU credited the doctoral student with web and graphic design. With David Kurt Herold, he then coauthored a 2015 review of twenty years of Chinese Internet research.^{17, 18} Helping maintain a conference and questioning a field’s inherited agendas were different kinds of participation in the same emerging profession.

His 2013 research on trolling illustrates what the agenda could miss. On Weibo, Douban, QQ and Baidu Tieba, he examined local terms such as penzi, sprayer, and diaoyu, fishing. Users’ distinctions among abuse, provocation, deception and play changed the apparent object.¹⁹ Translation was part of the analysis. Applying an English term first and treating everything as its example could erase the knowledge he had come to learn.

Fieldwork brought another complication. Traveling through eight Chinese cities, staying with friends, transcribing conversations and following communication between platforms did not produce a naturally bounded field, an unambiguous participant position or unquestioned authorial authority. Those limits became the argument of Three Lies of Digital Ethnography, first an essay in 2018 and a journal article in 2020.²⁰ With Crystal Abidin, he co-edited Private Messages from the Field, making discomfort and ethical uncertainty available for methodological discussion.²¹

One friend’s movement between platforms gives the problem a recognizable scale. After losing VPN access, the friend complained in Chinese on WeChat; after restoring it, he posted an English expression of purposelessness on Facebook. Incidental conversation also proved more revealing than formal interviews’ generalities.²² The encounter resisted a geography organized solely around a restricted Chinese Internet and a free global one. His postdigital inquiry followed technologies interwoven with ordinary life, where an administrative or technical boundary did not supply the whole social explanation.

What images make possible, and what collecting changes

De Seta’s Dajiangyou Tumblr was already an experiment in publishing research in progress in 2014. Searchable images initiated interviews and group discussion; visitors sometimes contributed their own. An earlier version had been suspended after an image permitted on Weibo was prohibited on Tumblr. Rebuilding meant revising his assumption that content could pass unchanged between platforms. The archive’s circulation also brought him into contact with artist Michelle Proksell.²³

Collecting therefore did more than accumulate examples. It generated encounters, while moderation intervened in what he could circulate. A research tool had become part of the situation he was studying.

His work on biaoqing pursued similarly specific operations. Five years of hands-on QQ, Weibo and WeChat use, with Chinese urban fieldwork in 2013–16, supported a distinction among ASCII emoticons, standardized emoji, commercial stickers and custom pictures. QQ capture shortcuts and favorites, or WeChat’s incorporation of received pictures into personal collections, helped explain circulation.²⁴ Meaning depended on how someone obtained an image and whom they sent it to. The pictures were not simply a universal dictionary of feelings.

Everyday creativity also included aggression. In Wenming bu Wenming, he followed disputes over civility through platforms and personal networks in 2015–16, including Di Ba’s mobilization on Tsai Ing-wen’s Facebook page and a three-hour Skype interview with Christoph Rehage in January 2016. Patriotic legitimacy could be used to authorize abusive behavior.²⁵ Attending to users’ own distinctions did not require celebrating everything they did. Proper conduct was itself learned, enforced and reversed through encounters.

Proksell brought another practice to the shared material. Their 2015 article on zipai, photographing oneself, connected WeChat pictures to Chinese art and photography. An image made during air pollution could express care and bodily exposure, complicating a simple opposition between narcissism and resistance.²⁶ Interpretation needed the situation around the picture, while artistic presentation raised questions about what would happen when that situation was removed.

Their Cabinet of Moments archive began in early 2014, with Proksell in Beijing and de Seta in Hong Kong. Proksell did much of the collecting through WeChat’s People Nearby and recent publicly available Moments. By October 2017 they had more than 40,000 files from eleven countries. Folders, collaborative analysis, online compositions and exhibitions gave the images successive forms. Portable Domains let visitors browse on smartphones and contribute images back.²⁷

Returning an image to circulation could invite dialogue, but did not restore its original context or settle the ethics of collection. Their collaboration kept aestheticization and exoticism in view. The means of presenting knowledge had consequences for the people represented; artistic ingenuity could make that problem more visible without solving it.

Taiwan: rooms, listening and permission to remain

An Academia Sinica postdoctoral fellowship in Taipei from 2015 to 2017 overlapped the completion of the doctorate.²⁸ Taiwan gave de Seta another set of artistic and anthropological relationships, alongside opportunities to examine how a place could be represented through sound.

During a four-week 2016 residency in Xianqiao on Shanghai’s Chongming Island, televisions, scooters, amplified music, recyclers, construction and tourism complicated the expected rural soundscape. A storm arrived when he had no recorder, leaving a written account. He moved from long takes to short, directed vignettes, edited and mixed around temporal practices.²⁹ The missed recording made a condition visible: what a locality sounds like in a work depends on someone’s equipment, position and decisions about what to retain.

Against Sonic Naturalism, written in 2018, pursued the question through his recording practice, begun with dictaphones and tapes in 2006, and Taipei’s KINO-EAR festival. David Toop and Steven Feld helped him challenge promises of transparent sonic access. Microphones, editing, playback, the recordist and listeners’ expectations all helped constitute the encounter.³⁰ The point made recording more demanding, rather than making it useless. A convincing work needed to acknowledge the procedures through which someone could hear it.

Sensefield, which he curated in October 2017 with director P. Kerim Friedman alongside Taiwan’s ethnographic film festival, required a different listening and viewing infrastructure. Submissions needed multiple screens, four-channel sound or tactile interaction that a cinema could not offer. Two Bopiliao storefronts accommodated works by Lucien Castaing-Taylor, Ernst Karel and Véréna Paravel, Miyarrka Media, DJ Hatfield and Rahic Talif, and Yannick Dauby and Wan-Shuen Tsai. De Seta’s Nangang, Nangang used local consumer devices and recordings.³¹

The National Culture and Arts Foundation supplied NT$110,000 to the Taiwan Association of Visual Ethnography for the project. Free admission, continuous playback, bilingual labels and subtitles, and staff explanation brought it into encounters with passersby and residents.³² The curators also acknowledged that audiovisual forms still predominated despite their multisensory ambitions.³³ Changing the room altered what was possible without automatically fulfilling the curatorial proposition.

A 2019 Taipei Artist Village residency supported Listening/Accidents, shown in Nonverbal alongside Masayuki Harada and Shih Yun-yu. Its accidental sound art and Modes of Listening used cheap, repurposed playback equipment to investigate different listening habits; the institutional record names him 胡子哥.³⁴ The April release 19:16 (Accidental Listening) converted the four-channel installation into stereo: January–March smartphone recordings of Taipei passed through four consumer devices. Residency support and name-your-price distribution helped the work circulate beyond the exhibition.³⁵

These arrangements developed across an insecure employment transition. Visa restrictions forced him to leave Taiwan after the fellowship and freelance work. From Italy he applied for an Employment Gold Card in summer 2018; in 2019 he described himself as freelancing, between jobs and starting his career. Independent work permission offered a measure of stability.³⁶ A transnational practice could be interrupted by the conditions of legal residence, even where relationships and work remained.

Music had material arrangements of its own. His Monstres par Excès label operated from 2007 to 2017, producing small editions and reinvesting earnings in subsequent releases.³⁷ His later Taiwan music study drew on three years of participation to examine how musicians, organizers and audiences made a scene recognizable and connected it to other scales.³⁸ These were economies of making and circulating music, alongside his academic and freelance work.

The 2020 Naturalismo album Elegie holds another history. He developed its pieces over roughly six years, looking back to the deaths of Zev Asher, Aube and Zbigniew Karkowski in 2013. De Seta had met Asher in China in 2007, learned aesthetically from Aube’s minimal materials and repetition, and met Karkowski shortly before his death. Moa Pillar mastered the release; Ivan Zoloto designed it. Hand duplication in Barcelona and Risograph production in Antwerp made the cassette edition.³⁹

The album’s memories, techniques and credits describe an artistic obligation with its own duration. Grief and aesthetic apprenticeship help explain why a piece is made; the production chain explains how it reaches someone else. Neither requires treating the recording as an ethnographic illustration. This is another way relationships persist through his work.

A familiar application becomes an infrastructural question

The inquiry into everyday media widened without leaving the initial encounters behind. With Jean-Christophe Plantin, de Seta examined WeChat’s transformation into infrastructure in 2019, connecting technical documentation, business reporting, observation and interviews. Payment and access to services made users dependent on an application that remained an intimate communicative space; company development met techno-nationalist conditions.⁴⁰ What a friend did with an app could lead toward a political economy of systems.

A 2023 QR-code article followed Chinese field notes from 2014–18 through drinks, karaoke bills, concert tickets and airport displays directing passengers to mapping apps. Patents, technical documents, advertising and professional interviews reconstructed another itinerary: US barcodes, Japanese QR development and incorporation into Chinese platforms.⁴¹ A mundane interface had a transnational history. An ordinary payment could reveal standards and institutional choices that neither a national success story nor a purely local description would capture.

Editing a 2023 Global Media and China issue assembled nine studies around networks, systems and standards. His introduction used Huawei’s C&C08 telephone switch, whose rural installations required protection against rodents and unstable power, to bring infrastructural ambition down to material conditions.⁴² National modernization and censorship matter, but they do not explain every cabinet, maintenance task or domestic use. Editing gave him a way to bring other researchers’ scales of inquiry into the same discussion.

He was already mapping China’s AI industry in a February 2019 interview with Nicolas Nova. Post-AlphaGo enthusiasm depended on venture capital, public investment and procurement. Beijing, Shanghai and Shenzhen had distinct roles; data processing and annotation extended beyond spectacular laboratories. He questioned how many companies could survive if support receded.⁴³ This was an inquiry into the distribution of work and resources at a particular moment, rather than an industry forecast established by subsequent success.

His 2020 argument about sinofuturism revisited the expectation that had helped take him into Chinese study. Projecting futurity onto China could deny Chinese people a shared present, as older orientalism had placed them outside historical time. Johannes Fabian’s coevalness was a central resource.⁴⁴ An early enthusiasm had become a problem requiring criticism. The correction mattered to how he interpreted technological promises, including the ones now made by companies.

Bergen: changing the evidence, sharing the work

De Seta joined Jill Walker Rettberg’s Machine Vision in Everyday Life project in early 2020. He later described considerable freedom to pursue work shaped around his expertise. Bergen’s humanities faculty awarded him its 2021 young-researcher prize, presented in March 2022.⁴⁵ The ERC project ran from 2018 to 2024 with approximately €2 million in EU funding.⁴⁶ The appointment gave him a funded institutional setting for research with others.

The project put his Chinese media interests alongside games, narratives and digital art. In Machine Vision Situations, Marianne Gunderson, Ragnhild Solberg, Linda Kronman, de Seta and Rettberg traced distributed agency through 874 situations in 500 creative works. Credits assigned de Seta conceptualization, supervision and reviewing/editing; Gunderson led drafting as well as conceptual and methodological work; Rettberg contributed funding and administration.⁴⁷ A shared question required different kinds of labor. The collaboration enlarged the available evidence without making every participant responsible for every task.

Corporate access was harder to obtain, and the pandemic made it harder still, reshaping de Seta’s interest in imaginaries.⁴⁸ With Anya Shchetvina he examined twelve Chinese machine-vision companies’ websites between November 2021 and June 2022. Their approximately 170 pages of working notes and more than 300 images followed interfaces, videos and documentation in Chinese and English. The 2023 publication identified computational abstraction, human–machine coordination, smooth everyday life and dashboard realism as four visual registers.⁴⁹

Those materials established how an industry presented and imagined technology. They could not audit how the systems performed. The restriction was also an opportunity to ask how familiar global promotional conventions promised relationships among machines, people and environments. A changed route to the object required a changed account of what the findings warranted.

Language still offered another route. His 2021 AI Now entry on heikeji, black technology, examined associations with speculative power, opacity, danger, surprise, marketing and national ambition. Black referred to obscurity rather than the racial meaning of Black tech.⁵⁰ Fiction and everyday evaluation helped people position themselves toward products whose engineering they did not know, including skepticism about supposedly miraculous capacities.

Making a product work could disturb its promise more directly. For the 2024 Synthetic Ethnography article with Matti Pohjonen and Aleksi Knuutila, de Seta followed Bilibili face-swapping creators and discussions. He uploaded his own face singing Dame da Ne through the First-Order Motion Model on Google Colab; responses disclosed Google-access problems. A local FaceSwap attempt to replace Jackie Chan’s face with his own consumed hundreds of GPU hours and produced flawed results.⁵¹

The polished demonstration had hidden the work. Learning interfaces, obtaining computing resources and responding to others became part of the field device, a designed intervention that created a research encounter. Failure could disclose conditions of participation that watching a successful clip would miss. It did not, by itself, explain every model or user.

His Synthetic Probes experiment began in March 2023 with ModelScope Text2Video on Hugging Face. Six months yielded more than a thousand four-second videos through over eighty minimal English prompts about China. Repeated outputs and a stock-footage watermark supplied clues to WebVid training material. He made Latent China from selected generated clips, source footage, his authored narration spoken through a synthetic voice and generated music.⁵² Prompting, selection and editing remained human work inside the synthetic composition.

In 2025 he qualified what this procedure could discover: probing did not expose all of a model’s datasets or components. The watermark was a fortunate clue; code analysis, model archaeology and quantitative work were also needed.⁵³ This acknowledges a practical division of knowledge. A researcher can invent a revealing encounter while needing other expertise to establish what the machinery contains.

Folklore becomes a program, while art keeps other purposes

ALGOFOLK began in 2024 at Bergen’s Center for Digital Narrative and runs to 2028. Marianne Gunderson and Debarun Sarkar are postdoctoral researchers; Hanna Hellesø Lauvli and Yağmur Çisem Vik are doctoral researchers. Theory, global ethnography, narrative/media analysis and an archive form four work packages.⁵⁴ A question previously developed through de Seta’s own relationships is now distributed among investigators, sites and methods.

The Trond Mohn Starting Grant was valued at approximately €2 million including university cofunding.⁵⁵ Its recruitment pathway includes an academic position the candidate can apply for after the program, rather than guaranteed permanent employment.⁵⁶ The grant changes the means available for inquiry and his responsibilities to other people’s work.

The project name recovers the folklore category he had set aside to enter sociology. He subsequently developed it through cross-disciplinary reading and a 2020 chapter, rather than simply renaming the doctorate.⁵⁷ His 2026 Four Theses on Algorithmic Folklore proposes relations among older traditions, new technical niches, automation, vernacular creativity, literacy gaps and exchanges between expert and everyday knowledge.⁵⁸ The program’s distinctive question concerns people making sense of systems that can also generate and circulate their expressive material.

Long observation remains important. His 2026 Yilong Ma article follows the Chinese Elon Musk impersonator across years of platform circulation. A 2022 WeChat exchange with a professional face-swap creator used screenshots to compare visual details. Productive uncertainty, celebrity recognition and performed Chineseness mattered alongside whether the face was synthetic.⁵⁹ The inquiry followed what the uncertainty enabled, rather than requiring a single authenticity verdict before the social phenomenon could be studied.

He also joined Daniel Vogler and Adrian Rauchfleisch for a different form of evidence: a Swiss online experiment with 1,361 participants distinguishing short authentic and deepfake videos. Their 2025 study found no direct improvement from a brief visual-cue intervention, with effects related to experience and media literacy.⁶⁰ The result is limited to that intervention and material. Learning through his own experiments and helping others learn are related tasks, but their outcomes need separate evidence.

Teaching has meanwhile become part of his own occupation. Bergen records a bachelor-level machine-vision module in autumn 2020 and a master’s Key Theories in Digital Culture course in autumn 2021, alongside supervision, workshops and an algorithmic-folklore interview podcast.⁶¹ His publications page describes efforts to make writing accessible beyond paywalls.⁶² A research question has to become a teachable problem or an intelligible conversation as well as a journal article.

Artistic collaboration continues to pose questions in other forms. In November 2023, de Seta and Magnhild Øen Nordahl organized Replicator with Andreas Zissler, Funda Zeynep Ayguler and Hans Oscar Carlsson. A three-day process moved from generated text and images toward 3D modeling and physical production, supported through BEK and the EU-backed Re-Imagine Europe setting.⁶³ Making an object exposed transitions an image alone could conceal.

Mariella Otto’s 2025 Diviner [A] was a 160-page book, edition 200, supported by the Swedish Arts Grants Committee and serving as prompt and documentation for Diviner [B]. Their texts combined divination, entropy and technology with Markov chains, ChatGPT and chickens interacting with typing.⁶⁴ The devices made signs available for interpretation.

With Luigi Monteanni, de Seta’s 2025 Supernaturalist Listening approached another authority of sound through criticism and speculative composition. Its apocryphal future broadcast and spirit-box encounter belong to the text’s declared fiction.⁶⁵ The question connects to his earlier recording work: what does a listening procedure authorize someone to say they encountered?

De Seta has said creative work can express things a conventional paper cannot, and that he prefers making it to remaining exclusively in high-level theory.⁶⁶ The preference does not abolish the distinctions among an album, an installation, a model experiment and an argument. Each creates a different encounter and owes its audience a different account of what it is doing.

Learning from the arrangements people make

In a 2025 essay with Friedman, de Seta argued that algorithms’ rules arise through people, companies and institutions, and that opacity can serve organizational convenience. They called for users to help determine the conditions governing their feeds.⁶⁷ The proposal returns responsibility from an imagined autonomous machine to the arrangements through which it operates.

His career has made those arrangements tangible at several scales. A language program and musicians’ introductions opened a field; friendships complicated its boundaries; artists and curators changed how evidence could encounter an audience; residence permission determined whether work could continue in a place. A funded team now gives some questions a more durable institutional form, while also creating new responsibilities and dependencies.

The continuity lies in a willingness to let participation change the question. China as a promised future becomes a critique of withheld coevalness. A field recording’s apparent transparency becomes an inquiry into equipment and listening. An AI tool’s apparent ease becomes hours of failed work and a methodological limit. De Seta’s practice is strongest where it carries the specific conditions of those revisions into the account, while allowing music and other people’s knowledge to keep purposes that his research does not exhaust.

Notes and sources

Sources were checked in October 2026. Interviews retain their retrospective perspective; formal records distinguish appointments, thesis submission and degree award. The Matter of Imagination page gives conflicting 2024/2025 publication dates, recorded below. Audio and video works are discussed through published descriptions and credits, without claiming unavailable recordings were watched or heard. Full-text and access limits are noted where material.