Intelligence artificielle : merci pour l’aide précieuse d’un collègue qui a fait un PLS (Plain Lay Summary)

Le 14 novembre 2023, j’ai analysé un article et terminé le billet par « Cet article mériterait d’avoir un ‘Plain lay summary‘ pour communiquer plus largement les résultats. » Il s’agissait d’un billet dont le titre était « Est-ce que la plupart des résultats de recherche sont faux ? OUI, OUI mais personne ne s’inquiète ».

Un PLS est plus facile à lire que le résumé de l’article !!!

Je dois remercier Yves Goulnik qui en lisant ce billet a demandé à l’intelligence artificielle de faire un PLS ! Il m’a confié son résultat et je suis séduit. C’est très intéressant car il m’a aussi communiqué le prompt. (Le prompt c’est l’instruction rédigée en langage naturel — français ou anglais.. — et non en code informatique et qui sert à commander les IA génératives). Voici :

 

Summary of the Article: “Are most published research findings false? Trends in statistical power, publication selection bias, and the false discovery rate in psychology (1975–2017)”

Study Title: “Are most published research findings false? Trends in statistical power, publication selection bias, and the false discovery rate in psychology (1975–2017)”

Researcher: Andreas Schneck

Objective: The study investigates the reliability of scientific findings reported as statistically significant in psychology research. It examines statistical power, publication bias, p-hacking, and the false discovery rate.

Context: The replicability crisis and cases of fraud have raised concerns about the trustworthiness of scientific results. Incorrect findings can lead to misguided policy decisions and harm.

Methods:

• Analyzed 35,515 psychology papers (published 1975–2017) with 487,996 test values.
• Explored statistical power, publication bias, and p-hacking.
• Estimated the false discovery rate.

Key Findings:

1. Statistical Power:
   • Power was lower than the recommended 80%, except for large effects.
   • Power increased slightly over time.
2. Publication Bias and P-Hacking:
   • Substantial publication bias and p-hacking were observed.
3. False Discovery Rate:
   • Estimated 17.7% false discoveries among significant results.
   • Assuming 50% true hypotheses and p-hacking as the main mechanism.

Implications:

• Research results may suffer from low statistical power and bias.
• Many significant results likely contain substantial findings, not just artifacts.

Significance:

• Raises awareness about research practices.
• Calls for improved methods and transparency.

Further Research:

• Investigate alternative scenarios.
• Address limitations in assumptions.

In summary, this study sheds light on the challenges in psychological research and emphasizes the need for rigorous practices to enhance the reliability of scientific findings.

Prompt écrit par Yves Goulnik  (Merci à Yves)

Make a summary of the article « Are most published research findings false? Trends in statistical power, publication selection bias, and the false discovery rate in psychology (1975–2017) » available at the following URL: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0292717#sec013

– Concentrate on the following sections: Abstract, Introduction, Data and methods, Results, Robustness, Central limitation
– Be accurate, clear and concise
– Do not assume any prior knowledge
– Use words that are appropriate for a reader with basic college education
– Use short sentences (up to 20 words) and short paragraphs (up to 3 sentences)
– Use neutral languageConsider general information about the research such as:
– Study title
– Who carried out the research? (including details of sponsor, funding and any competing interests)
– What public involvement there was in the study (how many people, what their relevant lived experience was, and what they did)
– Where and when the study took place
– Why was the research needed?
– What were the main questions studied?
– How can this study help patients and researchers?
– Details of any further research plannedThe summary should be around 500 words long.