Human Biases Reflected As Computer Biases
Bias: prejudice in favor of or against a thing, person, or group
What is “computer bias”?
The existence of prejudiced outcomes in the decisions or predictions made by computer systems/algorithms. Bias can be implemented into algorithms because of human biases being intentionally inserted or because of the data utilized being biased.
Explicit Data: Information directly provided by user.
Implicit Data: Infomration that can inferred from explicit data.
Based on either explicit/implicit data that has been used to train an algorithm, whether intentionally introduced or during the process of training data generation, bias can be created.
A notable example of this is seen in Netflix, where there are is a human factor that drives bias: Netflix exclusives are placed ahead (a show that is a Netflix exclusive means that users will be more likely to stay with Netflix). The bias in this case is Netflix’s prioritzation towards Netflix-produced shows.
Hack 1:
What is another example of a human bias being implemented into an algorithm?
Answer: Picture brightening at night works better for white people because the algorithms were trained more on them.
Reducing Bias
Programmers should take action to reduce bias in algorithms used for computing innovations as a way of combating existing human biases. Softwares need to be unbiased, consider all everything, and reject human bias.
Things to consider when developing programs:
- What are potential sources of bias?
- Is your program enhancing or intentionally excluding?
- Are you receiving feedback from a widespread group of people?
- How could people who differ from you use your developments?
Hack 2:
What is another way a programmer can reduce bias in their softwares?
Answer: Establish a clear process to test and give information to the software.
Types of Bias in Software Development
Biases can be embedded at all levels of software development.
It can be intentional or unintentional. Some software development are made for a certain market and ensure that people of certain places or demographics can use them easily. However, this doesn’t mean that they are trying to exclude.
Examples:
Intentional:
- Games could be geared towards a certain age range (Talking Tom vs Valorant)
- Game concepts
- Music
- Visuals
- WeChat and KakaoTalk
- Almost everyone in China uses WeChat
- KakaoTalk is the Korean version
Unintentional:
- Social media, Facebook vs. instagram
Hack 3:
What are some other examples of intentional and/or unintentional bias in innovations (games, social media, technology, etc.)?
Answer: Intentional was on Disney movies as it is more for children. Unintentional bias was social media images going to a specific person.
Homework
Question 1:
Define “computer bias” in your own words and explain how it can result from intentional or unintentional factors in software development. Give a brief example of this. Explain how programmers can actively work to reduce bias in their algorithms?
Answer: Computer bias is where software is specifically tailored to a specific group. It can result from intentional factors given by developers or be accidental. An example of this is Disney movies are made specifically for children. Programmers can actively work to reduce bias in the algorithms by using equal sample sizes of all distinct groups.
Question 2:
Briefly describe the two types of bias in software development and provide examples from the gaming industry and social media platforms. How might biases in software design affect user engagement and experiences?
Answer: Intentional bias is tailored to age - games like fortnite are specifically for older teens while games like paw patrol are for toddlers and younger kids. Unintentional bias is like some camera programs work better for specific skin tones. This affects user engagement and experiences because some skin tones have worse experiences on these cameras.