Society / Social Change
Track social change, shifting values, public sentiment and cultural transformation through structured summaries built from curated sources.
Konrad Körding | Helping Human Scientists do Better Science @ Vision Weekend Puerto Rico 2026
Summary
Human scientists frequently struggle with poorly defined questions, which leads to research that lacks significant impact. Many scientists ask questions that do not yield meaningful answers, resulting in a plethora of studies that contribute little to the field. The introduction of AI into scientific research raises concerns about the accuracy and validity of the results produced.
AI systems primarily rely on existing data to generate predictions, but this data often contains biases and inaccuracies. The challenge lies in training AI to recognize what constitutes good science, as there are no clear metrics to define it. Many published papers, despite being flawed, receive acceptance, while valuable research may be rejected.
The emphasizes the need for AI to add friction rather than remove it from the scientific process. By encouraging critical thinking and thorough examination of ideas, AI can help scientists refine their research questions and methodologies. A proposed tool aims to assist researchers in identifying gaps in their logic and improving the quality of their work.
Concerns about the pressure to publish within the scientific community contribute to the proliferation of subpar research. The current publication system often rewards quantity over quality, leading to the acceptance of poorly conceived studies. This environment necessitates a reevaluation of how research is validated and shared.
Perspectives
Discussion on the role of AI in enhancing scientific research quality.
Pro-AI in Science
- Advocates for AI to add friction in scientific processes
- Encourages critical examination of research ideas
- Proposes tools to help scientists refine their logic
- Highlights the need for clear definitions of good science
- Calls for higher standards in the publication process
Skeptical of AI's Role
- Questions the effectiveness of AI in producing valid scientific insights
- Points out biases in existing data used for AI training
- Critiques the pressure to publish leading to poor research quality
- Expresses doubt about AIs ability to generate truly creative ideas
Neutral / Shared
- Acknowledges the existence of useful AI tools in science
- Recognizes the challenges of defining good science
Metrics
other
100 ways how people get it wrong ways
common failure modes in science
Identifying these failures can help improve scientific methodologies.
I have seen science fail in probably about 100 ways.
other
a massive checklist of how science can go right
defining good science
A checklist could standardize what constitutes good science.
We need a massive checklist of how science can go right.
other
the first drug candidate was produced by an AI drug candidate
AI's role in drug discovery
This highlights the potential of AI in pharmaceutical research.
the first drug candidate was produced by an AI.
Key entities
Timeline highlights
00:00–05:00
Human scientists often struggle with poorly defined questions, leading to research that lacks impact. The increasing use of AI in scientific research raises concerns about the accuracy and meaningfulness of the results produced.
- Human scientists often struggle with poorly defined questions, which can lead to research that lacks impact. This emphasizes the importance of improving question formulation in scientific studies
- Even with the right questions, scientists may reach incorrect conclusions due to flawed methodologies, raising doubts about the reliability of their findings
- The use of AI in scientific research is increasing, with new tools designed to assist scientists. However, there is skepticism regarding the accuracy and meaningfulness of the results produced by these AI systems
- AIs reliance on existing data for predictions can reinforce flawed research, making it difficult to differentiate between valuable and worthless studies
- Current AI applications in science tend to focus on streamlining the research process, which may not enhance the quality of published papers. There is a need for AI to create challenges that promote critical thinking and thorough evaluation
- A proposed solution is to develop AI tools that encourage scientists to clarify their reasoning and identify gaps in their research. This approach aims to improve the quality and impact of scientific contributions
05:00–10:00
Human scientists often formulate vague questions, resulting in research that lacks significant impact. The increasing use of AI in scientific processes raises concerns about the accuracy and meaningfulness of the results produced.
- Human scientists often formulate vague questions, resulting in research that lacks significant impact, indicating a need for better question formulation in scientific inquiry
- While AI is being utilized to enhance scientific processes, it frequently produces inaccurate results, which can exacerbate existing issues in research methodologies
- The absence of clear metrics for defining good science complicates the effective training of AI, limiting its potential contributions to research
- The speaker suggests that AI should introduce challenges in the scientific process to promote critical evaluation of ideas and methodologies among researchers
- The pressure to publish can lead to an increase in low-quality research, as the academic system prioritizes quantity over the quality of findings
- A restructured research validation process that leverages AI could improve critical thinking in peer review, helping to filter out subpar research