A world of AI-assisted writing and reviewing might transform the nature of the scientific paper.

  • flossdaily@lemmy.world
    link
    fedilink
    English
    arrow-up
    5
    arrow-down
    1
    ·
    1 year ago

    The main probably with scientific publishing is that our threshold for statistical significance is way too low.

    If we allow the threshold to sit at a 1 percent chance that results of the study were random chance, it means that 1 percent of all publications at that level of certainly are going o mislead the public if the media reports on them. And with the volume of research published every day, that adds up to a LOT of misinformation.

    It’s not even bad science, it’s bad reporting and widespread scientific illiteracy. But neither of those are going away.

    • sndrtj@feddit.nl
      link
      fedilink
      English
      arrow-up
      3
      ·
      1 year ago

      Also certain fields (cough cough medicine) needs to consider more than just the p value. With any large sample size you’re almost guaranteed to find a “significant” result in some test, but the effect sizes are often so tiny to be basically meaningless.

    • Zeth0s@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      arrow-down
      1
      ·
      edit-2
      1 year ago

      Problem of science is corrupted funding, toxic environment, mafia-like organizations and practices, exploitation, widespread corruption.

      The current amount of bad science is due only to that. Even before reaching mainstream media, of which I don’t care. Unmanageable excess of meaningless published work is just a side effect of all above. Clearly it cannot change from inside, as current system selects only those who agree or compromise.

      A revolution must come from outside. Tools and platforms like arxiv, github, hugging face are already demonstrating that alternative way of working exists, better ways to spread science and facilitate collaboration, increasing quality. Unfortunately they do not currently represent a real alternative outside niche fields, were quality, reproducibility and speed of evolution are critical. But also alternative tools such these can alleviate a minimal part of the huge problems.

      I am honestly curious too see how “scientific” system will evolve, because it will. Because as it is now it is doomed to miserably continue falling down even further…