Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Analysing panel co-attendance in scientific conferences

Analysing panel co-attendance in scientific conferences

Presentation at the #Sunbelt2020 virtual session on 𝘽𝙞𝙥𝙖𝙧𝙩𝙞𝙩𝙚 𝙉𝙚𝙩𝙬𝙤𝙧𝙠𝙨 & 𝙋𝙧𝙤𝙟𝙚𝙘𝙩𝙞𝙤𝙣𝙨, #NETCONF research project



June 10, 2021


  1. Analysing panel co-attendance in scientific conferences A new avenue to

    explore academic sociality #Sunbelt2020 virtual sessions on 𝘽𝙞𝙥𝙖𝙧𝙩𝙞𝙩𝙚 𝙉𝙚𝙩𝙬𝙤𝙧𝙠𝙨 & 𝙋𝙧𝙤𝙟𝙚𝙘𝙩𝙞𝙤𝙣𝙨 Empirical Insights from Bipartite and Multipartite Networks July 15, 2020 François Briatte · ESPOL, Catholic U. of Lille, France Marion Maisonobe · Géocités, CNRS, Paris, France
  2. None
  3. Background (network science + science studies) • Research on scientific

    collaboration and networks mainly relies on co-authorship data • Many other types of interactions contribute to the circulation of ideas between scholars and to the emergence of scientific groups • Among them, social links derived from participation to conferences and panel co-attendance • Interesting to explore, with conference programmes and lists of participants often openly accessible on the Web
  4. • Panel co-attendance in two scientific conferences about green chemistry

    (ISGC) and political science (AFSP) • Conference structure (sessions, topics, etc.) and its evolution can inform specialty/discipline dynamics • Panel co-attendance used as a proxy for knowledge circulation between participants/places • Common methodology applied to both conferences, through several conference years T (ISGC) = 6 years, T (AFSP) = 10 years (every odd year) Goals of the project
  5. Project summary • NETCONF project, 2020–21 (funded by URFIST GIS)

    • Participants: CRIEF (EA, Univ. Poitiers), ESPOL-LAB (Univ. Catho. Lille), Géographie-cités (UMR CNRS, Univ. Paris 1, Univ. de Paris, EHESS), CESSP (UMR CNRS, Univ. Paris 1), ELICO (EA, Univ. Lyon), INCREASE (FR CNRS) • Intended outputs: (1) clean network data, and measures of (2) internationalization and (3) geographic span for both studied specialties/disciplines • Groundwork for inclusion of additional conferences (possibly from different fields)
  6. None
  7. None
  8. Comparable properties • Both conferences count ~ 700–1,000 participants i,

    with most attending a single panel j • Attending several years is frequent but not the modal behaviour (return rate ~ 25% for ISGC, ~ 30% for AFSP) • ISGC (green chemistry, organised in La Rochelle) is much more internationalized (50–60%) than AFSP (French or Francophone political science, itinerant, max. 20% intl.) • Yet both conferences attract participants from roughly the same number of countries (~ 60)
  9. Geographic ties based on panel co-attendance Francophone countries

  10. Each tie captures one panel co-attendance in the panel/participant bipartite

    graph. We used all ties > 1 (2+ panel co-attendance at the same conference) to extract the backbone of each graph. Universal method used, but perhaps edge distribution justifies another approach? Edge weight distributions of one-mode projections 1 tie 2 ties 3 ties 4 ties 5 ties ISGC 2015 17080 122 6 4 ISGC 2017 20444 230 26 4 2 ISGC 2019 18230 358 24 8 4 AFSP 2009 14478 230 8 AFSP 2011 9538 34 AFSP 2013 15364 100 2 AFSP 2015 17476 114 AFSP 2017 9934 52 AFSP 2019 15180 82
  11. None
  12. Next steps (suggestions welcome) • Weighting scheme of the (

    Participant i × Panel j ) bipartite adjacency matrix: 1/N i weights? • Additional data on participants (e.g. affiliations, publications, dissertation committees) — ongoing work • Characteristics (socio-demographic, geographic) of the nodes connected by the network backbone • Temporal analysis with TERGMs? (institution-/city-level homophily sustained through conference years; cohorts) • Additional data from similar conferences? see experimental takes at epsa2020 and statconf
  13. • ISGC conference data shared by its managing company •

    AFSP conference data scraped from the AFSP website see congres-afsp for the code and (preliminary) data • Graph visualizations performed with igraph (Csárdi), ggraph (Pedersen) and graphlayouts (Schoch) • Backbone extraction performed with backbone (Domagalski, Neal and Sagan) • References for final report on Zotero Sources
  14. Thank you for your attention Slides at frama.link/netconf-2020-sunbelt marion.maisonobe@cnrs.fr ·

    @GeoMaisonobe francois.briatte@sciencespo.fr · @phnk