quartz/content/notes/ethics.md
2022-04-06 20:35:03 +12:00

64 lines
1.8 KiB
Markdown

---
title: Ethics
draft: true
---
coded biases doco
# Ethics
## 1 Case studies
1. [facial recognition in US riots 2021-01-06](out/notes/facial-recognition-in-us-riots-2021-01-06.md)
2. [Anti govt protest china](out/notes/anti-govt-protest-china.md)
3. [How is safe enough for autonomous vehicles](out/notes/how-is-safe-enough-for-autonomous-vehicles.md)
### 1.1 Differences 1 vs 2
Govt vs vigilante
my judgements contain additionl context
e.g., pro-democratic vs anti
world contains vast differences
how systems of laws work
extent of civil liberties afforded to individuals
### 1.2 Discussion
When developing a technology you dont know what is could be used for
## 2 Ethical handling of data
- Data moves very quickly due to computerised systems
- privacy act 2020
- its unethical to ignore potential security problems
- df
## 3 Ethical handling of bias and errors, e.g., in AI
- large datasets oftenb incdlude bias and errors
- to AI trained on these datasets with also be biased
- e.g., facial recognition trining overrepresenting white males
- ML algorithgms are often opqaue
- its not possible to understand how decisions are reached
- makes asessing suitability of AI for a use case difficult
- explainable AI
- attacks e.g., 'trapdoors' within ML training data
## 4 False or misleading claims
- pressure to release can lead to false claims
- are features fully tested
- need to assess risks of bias
- e.g., AWS uptime information
- rumoured that service status colour is n management decision
-
## 5 Your responsibility
- dont stay silent
## 6 Professional reponsibilities
- comp science per se lacks profressional standards
- there are some prefessional bodies which encoede responsibilities
- ACM coc
- IEEE coc
- neither are specific to NZ
- Within NZ must consider treaty obligations