Are ChatGPT and Copilot Reliable for Health Education on Statistical Testing?
Abstract The introduction of Artificial Intelligence (AI) has revolutionized daily life and scientific research, with applications ranging from writing scientific articles to clinical assistance. However, the effectiveness of AI models like ChatGPT 3.5 by Open AI and Bing Copilot GPT-4 by Microsoft in explaining complex concepts such as statistical testing is a cause for concern. This study investigates the ability of these AI models to explain fundamental statistical concepts, such as P-values, confidence intervals, and surprisals, crucial to properly inform conclusions in scientific research and public health. Our results highlight significant misconceptions in both AI models’ understanding and teaching of inferential statistics. These deficiencies include the mixing of incompatible statistical approaches, the nullism fallacy, the dichotomization of (statistical) significance, the incorrect interpretation of statistical measures and concepts, and an overestimation of the role of p-values and confidence intervals. Additionally, both models lack knowledge of recent alternative statistical methods like S-values and S-intervals, showing biases similar to those present in traditional statistical approaches. Given the importance of accurate statistical understanding in various sectors and the widespread integration of AI in decision-making processes, urgent intervention by OpenAI and Microsoft is necessary to update their platform databases. It is essential to align AI knowledge with the latest developments in scientific research to ensure the reliability of generated results. Collaboration with organizations such as the American Statistical Association is recommended to facilitate this process. In conclusion, this scenario underscores the need for immediate corrective action by the developing companies of such platforms. Indeed, only through continuous updates and improvements can we ensure that AI can contribute positively to scientific and technological progress..
| Medienart: |
|
|---|
Preprint| Erscheinungsjahr: |
2024 |
|---|---|
| Erschienen: |
2024 |
| Enthalten in: |
bioRxiv.org - (2024) vom: 14. März Zur Gesamtaufnahme - year:2024 |
|---|
| Sprache: |
Englisch |
|---|
| Beteiligte Personen: |
Rovetta, Alessandro [VerfasserIn] |
|---|
| Links: |
Volltext [kostenfrei] |
|---|
| Themen: |
|---|
| doi: |
10.1101/2024.03.08.24304007 |
|---|
| funding: |
|
|---|---|
| Förderinstitution / Projekttitel: |
|
| PPN (Katalog-ID): |
XBI042884225 |
|---|
| LEADER | 01000caa a22002652 4500 | ||
|---|---|---|---|
| 001 | XBI042884225 | ||
| 003 | DE-627 | ||
| 005 | 20240315090611.0 | ||
| 007 | cr uuu---uuuuu | ||
| 008 | 240313s2024 xx |||||o 00| ||eng c | ||
| 024 | 7 | |a 10.1101/2024.03.08.24304007 |2 doi | |
| 035 | |a (DE-627)XBI042884225 | ||
| 035 | |a (biorXiv)10.1101/2024.03.08.24304007 | ||
| 040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
| 041 | |a eng | ||
| 100 | 1 | |a Rovetta, Alessandro |e verfasserin |0 (orcid)0000-0002-4634-279X |4 aut | |
| 245 | 1 | 0 | |a Are ChatGPT and Copilot Reliable for Health Education on Statistical Testing? |
| 264 | 1 | |c 2024 | |
| 336 | |a Text |b txt |2 rdacontent | ||
| 337 | |a Computermedien |b c |2 rdamedia | ||
| 338 | |a Online-Ressource |b cr |2 rdacarrier | ||
| 520 | |a Abstract The introduction of Artificial Intelligence (AI) has revolutionized daily life and scientific research, with applications ranging from writing scientific articles to clinical assistance. However, the effectiveness of AI models like ChatGPT 3.5 by Open AI and Bing Copilot GPT-4 by Microsoft in explaining complex concepts such as statistical testing is a cause for concern. This study investigates the ability of these AI models to explain fundamental statistical concepts, such as P-values, confidence intervals, and surprisals, crucial to properly inform conclusions in scientific research and public health. Our results highlight significant misconceptions in both AI models’ understanding and teaching of inferential statistics. These deficiencies include the mixing of incompatible statistical approaches, the nullism fallacy, the dichotomization of (statistical) significance, the incorrect interpretation of statistical measures and concepts, and an overestimation of the role of p-values and confidence intervals. Additionally, both models lack knowledge of recent alternative statistical methods like S-values and S-intervals, showing biases similar to those present in traditional statistical approaches. Given the importance of accurate statistical understanding in various sectors and the widespread integration of AI in decision-making processes, urgent intervention by OpenAI and Microsoft is necessary to update their platform databases. It is essential to align AI knowledge with the latest developments in scientific research to ensure the reliability of generated results. Collaboration with organizations such as the American Statistical Association is recommended to facilitate this process. In conclusion, this scenario underscores the need for immediate corrective action by the developing companies of such platforms. Indeed, only through continuous updates and improvements can we ensure that AI can contribute positively to scientific and technological progress. | ||
| 650 | 4 | |a Biology |7 (dpeaa)DE-84 | |
| 650 | 4 | |a 570 |7 (dpeaa)DE-84 | |
| 700 | 1 | |a Mansournia, Mohammad Ali |0 (orcid)0000-0003-3343-2718 |4 aut | |
| 773 | 0 | 8 | |i Enthalten in |t bioRxiv.org |g (2024) vom: 14. März |
| 773 | 1 | 8 | |g year:2024 |g day:14 |g month:03 |
| 856 | 4 | 0 | |u http://dx.doi.org/10.1101/2024.03.08.24304007 |z kostenfrei |3 Volltext |
| 912 | |a GBV_XBI | ||
| 951 | |a AR | ||
| 952 | |j 2024 |b 14 |c 03 | ||