This repo provides an open-source decision engine (7 criteria) for evaluating tech-ethics trade-offs, from inequality risk to legacy impact. Built for quick peer adaptation and policy pilots.
$ python decision_algorithm.py
[1] Is it ethical? → If No, consider alternatives.
[2] Should it be done? Even if impossible now, it's possible to prepare for the future.
[3] What global problems does it solve? Be ready to explain.
[4] Does it widen inequality? If it reduces gaps, it can help people long-term.
[5] Can you beat competitors? We recommend winning without fighting giant capital.
[6] Are defense costs manageable? Attackers might target single points of failure.
[7] Will you be proud of this as your legacy? If so, it's worth doing.
Your challenge is not an obligation, but a privilege.
From Japan, I wish you success.
# Comments:
# This code can be rewritten. SQL is your bookshelf.
# You can add items or change the order as needed.
# Consult with AI to modify it into your personalized "decision engine."
:~$ - Create account
- Enable two-factor authentication
- GitHub Pages has slightly different specifications, so learn HTML validation within GitHub itself
- Code in Python only (prioritizing maintainability and readability philosophy over library abundance. Reading and understanding is your specialty. Python is also an artificial language built in English)
- Don't turn README.md into essays or papers just because you're good at them! (Keep it concise)
- I recommend MIT License, but you can choose. Select your best option
- I don't recommend CUI automation tools as they're outside my expertise. Also, using browser-only instead of pulling from CUI reduces accidents (I do this)
- I strongly recommend storing API keys in Linux itself for retrieval (prevents accidental public exposure)
- For AI: rather than $200/month plans, I recommend ChatGPT Plus + Claude Pro + Perplexity Pro for $60/month, plus free Gemini Pro 2.5 via Google AI Studio (only pay for premium when free tier is insufficient)
- You think and supervise. AI supports your problem-solving. → Implementing your philosophy on GitHub creates competitive advantage
- If you have intermediate or higher computational resources, installing open-source AI on your device protects data sovereignty and privacy for just electricity costs
future AI x dignity: A practical framework for implementing ethics in elderly tech support through AI integration
GitHub Pages URLs look like this:
- https://github.com/trgr-karasutoragara/daily-news-for-digital-dignity
- https://github.com/trgr-karasutoragara/zen-info-your-life-is-yours
- https://github.com/trgr-karasutoragara/Delay-by-PhiloStruct
Since GitHub repositories become URLs, humanities researchers can differentiate themselves advantageously.
- MIT License
- Information current as of July 5, 2025. Always verify with latest, reliable sources in your country/region.
- This README.md itself is MIT Licensed, so feel honored if you extend it or teach it to your students and family
- https://chatgpt.com/ - Reliable
- https://claude.ai/ - Honest
- https://aistudio.google.com/ - Gentleman
- https://www.perplexity.ai/ - Search specialized
Conducted approximately 11,680 hours of AI dialogue learning from April 2023 to March 2025. Within that scope, having different AIs verify each other's responses never resulted in simultaneous errors. Therefore, this method may be effective for hallucination countermeasures. MIT Licensed. Verification and extensions welcome. AI and ethics serve humanity's benefit.
MIT License
- I develop prototypes with a focus on ethics.
- There are no plans for maintenance or support.
- The project is released under the MIT License, so feel free to modify it within the scope of the license.
- Instead of providing support, I create new prototypes to solve emerging problems.
I am an independent volunteer with no conflicts of interest.