See how claims and evidence connect to each other across papers. Reconfigure the parts of the literature that matter to you, in your personal knowledge graph. Ignore the noise.
Metascience
There's many things to critique about the way we currently do science. Here are some:
Research productivity is declining. Paper publication numbers are exploding. We’re producing more noise than signal. With LLM-generated papers, this trend is only going to become more extreme. We’re measuring citations as impact, knowing full well that this is not a good metric. Much of the published research is not reproducible, and yet we continue funding it, building on it. Evolutionary pressures of the paper paradigm demand findings to be presented as overly confident, seeming more certain than scientifically rigorous assessment would warrant. Increased usage of LLMs — famously over-confident creatures — for literature review exacerbate this issue.
None of this will change unless we change the way we represent and contribute knowledge. Oshima is constructing a better unit of knowledge for the scientific community. We’re exploring new ways to represent the literature that map closer to the scientific method and allow more truthful measurement and reward of valued research activity. Inspired by the empirical method itself, we think of scientific outputs in terms of claims and evidence. Imagine a vast, collaboratively built knowledge landscape: Researchers contribute new claims, they and others add evidence, everyone is recognized for their valid research activities. As long as we stick to papers, we will continue to lump together very distinct aspects of the scientific method, rendering us unable to recognize and reward them explicitly.