Athabasca University | AU Student/Staff Login | Invited Guest Login
The DIKW hierarchy as interpeted by ChatGPT
The DIKW hierarchy
Many web-based resources discuss DIKW, Li & Duan's [1] paper explored the usage of DIKW by ChatGPT [2]. The process, as interpreted by [2], provides valuable insight into the extraction and interpretation of Data, Information, Knowledge and Wisdom from the initial statements. The accuracies of the extraction may be questionable and the critique is left up to the reader.
The following is ChatGPT [2] output produced using the first example (3.1) from [1].
|
Layer |
Data |
Information |
Knowledge |
Wisdom |
Purpose |
|
Data |
17-year-old, 180 catties, obese, Tom weighs 185 catties, decreased appetite, weight loss |
|
|
|
|
|
Information |
Weighing more than 180 catties at 17 years old is considered obese, Tom is obese |
Tom worries about his weight, saw news about obesity, received a weight loss flyer |
Tom's decreased appetite may be related to his weight, weight loss class |
Tom may benefit from seeking medical help for his weight concerns |
The weight loss training class requires students to not be obese |
|
Knowledge |
Obesity increases the risk of heart disease |
Obesity can cause decreased appetite, medical help may be needed |
Weight loss training may be beneficial for obese individuals |
Seeking medical help and enrolling in a weight loss program can benefit Tom |
The weight loss training class is not suitable for Tom as he is considered obese |
|
Wisdom |
Maintaining a healthy weight is important for reducing the risk of heart disease |
Tom should seek medical help for his weight concerns and consider a weight loss program |
Tom may benefit from attending a weight loss class |
Making informed choices about weight management can improve health outcomes |
Tom should prioritize seeking medical help and weight loss for his overall health |
To visualize the relationship between the layers, we can use a pyramid-shaped diagram with data at the bottom and purpose at the top. The layers build on each other, with data being transformed into information, then knowledge, wisdom, and ultimately purpose.
```
Purpose
^
|
Wisdom
|
|
Knowledge
|
|
Information
|
|
Data
```
To create a graph for each layer, we can use a bar chart to show the quantity or frequency of data, a line graph to show trends or patterns in information, a scatter plot to show correlations in knowledge, a heat map to show relationships between different pieces of wisdom, and a pie chart to show the distribution of purpose.
For example, a graph of the data layer might show the number of teenagers who weigh more than 180 catties by age, while a graph of the information layer might show the prevalence of heart disease among obese individuals. A graph of the knowledge layer might show the relationship between obesity and other health conditions, while a graph of the wisdom layer might show the most effective strategies for weight management. Finally, a graph of the purpose layer might show the proportion of individuals who successfully lose weight after enrolling in a weight loss program.
Reference:
[1] Li, Y., & Duan, Y. (2023). The Wisdom of Artificial General Intelligence: Experiments with GPT-4 for DIKWP. arXiv preprint.
[2] ChatGPT, personal communication, May 4, 2023
- Two stories about learning to be human from a machine
January 12, 2023 - 5:22pm
Jon Dron - View of Speculative Futures on ChatGPT and Generative Artificial Intelligence (AI): A Collective...
March 13, 2023 - 12:49pm
Jon Dron
Comments
Critique of ChatGPT DIKW output
Concerning DIKW and the data from example (3.1) [1]. Without human processing, the statements can be considered a set of singly-linked lists of symbols, therefore, the data [3][4].
The data produces the following information after an initial conversion [3][4]. Further processing of the logical statements is possible.
The information establishes that Tom is obese, has a risk of heart disease, worries about weight, has a decreased appetite, knows about obesity, knows about weight loss training, and wants to go to the hospital. The information row from the ChatGPT [2] establishes the information.
Does ChatGPT provide knowledge from the information?
The knowledge row [2] provides information in the following statements:
“Obesity increases the risk of heart disease”
“The weight loss training class is not suitable for Tom as he is considered obese”
These statements can be derived from processing the initial information and are not knowledge [6][7]. The columns do provide actionable knowledge.
ChatGPT did provide some words of wisdom, "knowing the right things to do” [7] by providing the statements “Maintaining a healthy weight is important for reducing the risk of heart disease” and “Making informed choices about weight management can improve health outcomes” [2].
References:
[1] Li, Y., & Duan, Y. (2023). The Wisdom of Artificial General Intelligence: Experiments with GPT-4 for DIKWP. arXiv preprint.
[2] ChatGPT, personal communication, May 4, 2023
[3] Ackoff, R. L., "From Data to Wisdom", Journal of Applies Systems Analysis, Volume 16, 1989 p 3-9
[4] http://www.systems-thinking.org/dikw/dikw.htm
[5] http://www.spreadingscience.com/our-approach/diffusion-of-innovations-in-a-community/1-the-dikw-model-of-innovation/
[6] https://en.wikipedia.org/wiki/Tacit_knowledge
[7] https://en.wikipedia.org/wiki/DIKW
I love this! Excellent use of the technology and wonderful philosophical questions, Eugene.