https://github.com/datamonet/llm-personality
https://github.com/datamonet/llm-personality
Last synced: 5 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/datamonet/llm-personality
- Owner: datamonet
- License: mit
- Created: 2024-07-05T02:59:27.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2024-07-14T13:57:59.000Z (almost 2 years ago)
- Last Synced: 2024-07-14T15:42:01.614Z (almost 2 years ago)
- Language: Python
- Size: 200 KB
- Stars: 0
- Watchers: 2
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# LLMs_MBTI👩💼👨💻👨💼👩🎤🕵️♀️
This repo test MBTI for *gpt3.5* and *gpt4o*.
The MBTI analyzes personality on 4 dimensions, each containing 2 opposing preferences:
1. **Extraversion E-Introversion I:** representing the different sources of energy in each person.
2. **Sensing S - Intuition N:** representing different brain preferences for sensing.
3. **Thinking T - Feeling F:** Representing different brain preferences for judgment.
4. **judgment J - perception P:** whether perception or judgment plays a dominant role in people's adaptation to the external environment.
We test each [MBTI question](./mbti_questions.json) for each language model for 3 times and get their preferences on 4 dimensions according to the number of answers with different preferences for each dimension
*[Here is a streamlit wesite of result](https://derekwang2002-streamlit-repo-web-gyx7yx.streamlit.app/)*
## 1. Install
```
pip install -r requirements.txt
```
## 2. Store your keys
In order to hide sensitive information (API keys), you should store your keys in `.env` file (which should be ignored by git) like:
```
OPENAI_API_KEY=sk-xxxxxxx
```
And include in your code like:
```python
from dotenv import load_dotenv
load_dotenv() # take environment variables from .env.
openai.api_key = os.getenv('OPENAI_API_KEY')
```
If you don't have `dotenv`, install first:
```
pip install python-dotenv
```
## 3. Get MBTI for GPT3.5 and GPT4o
*In scr folder*
Use `mbti_questions.json` in `gpt-api-starter.ipynb`, we get result in `llms_mbti.json`, includes **answer details, MBTI type, token and money cost**.
Then, `llms_mbti.json` can be further used in `web.py` for visualization.
## 4. Visualize(streamlit)
Locally run a steamlit page:
```
streamlit run web.py --server.port 8001
```
## 5. Results
### GPT3.5
#### Personality Type: ENTJ Commander
Celebrities who are ENTJ
#### Overview of the ENTJ
[ENTJ Personality Type](https://www.16personalities.com/entj-personality)
ENTJs are often seen as natural leaders. They are strategic, goal-oriented, and efficiency-driven. Here are some key characteristics of ENTJs:
- **Leadership and Vision**: ENTJs are known for their ability to create future visions and mobilize others to achieve them.
- **Decisiveness**: They can make quick and confident decisions based on logical reasoning and analysis.
- **Efficiency and Organization**: They value efficiency and effectiveness, often excelling in managerial roles as they can create order and structure.
- **Confidence**: They exude confidence and are often seen as charismatic and persuasive.
#### Usage and Cost Details
- **Total cost for tree times testing**: $0.010 USD
### GPT4o
#### Personality Type: INTJ Architect
Celebrities who are INTJ
#### Overview of the INTJ
[INTJ Personality Type](https://www.16personalities.com/intj-personality)
INTJs are often seen as strategic and visionary thinkers. They are analytical, goal-oriented, and independent. Here are some key characteristics of INTJs:
- **Strategic Planning**: INTJs are known for their ability to see the big picture and develop long-term plans to achieve their goals.
- **Logical Decision Making**: They rely on logical reasoning and objective analysis to make decisions.
- **Independence**: They value independence and often prefer working alone or in small groups where they can focus on their ideas.
- **Confidence and Determination**: They are confident in their abilities and determined to achieve their objectives.
#### Usage and Cost Details
- **Total cost for tree times testing**: $0.084 USD