The need for Critical Thinking skills to navigate Fairness and WEIRDness in algorithms
- Sinead Quealy
- 5 days ago
- 2 min read
A few recent papers and articles caught our eye. Each consider bias, fairness and in the case of "Which humans?" paper, the WEIRD-ness of our interactions with AI. Here, WEIRD is Western, Educated, Industrialised, Rich & Democratic, and found that as the distance from the US increases, the GPT's "alignment with local human values declines."
The next piece puts us in mind of George Bernard Shaw's line, "the single biggest problem in communication is the illusion that it has taken place". In the paper, Panciroli and Rivoltella explore language, values and the concept of fairness - and unfairness. They consider that perhaps a way to look at AI is not as artificial intelligence, but artificial communication:
The first consideration, as we anticipated above, is in line with Elena Esposito’s (2022, p. 16) thesis that «machine learning programs realize not an artificial intelligence but a kind of artificial communication». Esposito points out that one can communicate with a machine if one emancipates oneself from the mentalistic meaning, which necessitates attributing intentionality to the interlocutor in order for one to be able to speak of communication, and places oneself in Luhmann’s viewpoint for whom «communication takes place not when someone says something, but when someone understands that someone has said something» (Ibi, p. 19).
In their conclusion, they suggest guidelines for many audiences and cohorts.
In essence, it is about creating the conditions for users to recognize the biases that may be present in models and to be able to take them into account when relying on them to make decisions. This awareness work can include the creation of guidelines, directories of best practices, and educational materials all aimed at enabling responsible use of applications. Development companies can also do their part by setting up tools and platforms that can facilitate customization and control.
Here again, we were reminded of the recent "It's Not the AI - It's Each of Us!" paper which sets out a suggested Ten Commandments for the Wise & Responsible Use of AI, broken down to sections on:
Purpose: Protect Humanity
Safeguards: Ensure Non-Negotiable Principles
Action: Proceed Purposefully, and
Reflection: Continuously Evaluate the Purpose-Impact Fit.
A common defence against the negative impacts of poor AI experiences is CRITICAL THINKING. Explicitly, Rivoltella's paper identifies as the paper's conclusion "the need for critical thinking for developers and users."
Each of these papers are worth your consideration and time.