Six Things a Diversity & Inclusion Chair Does
EACL 2024 is happening in Malta this week. What a great time to look into the role of the Diversity & Inclusion Chair at a conference and what they actually do.
ChatGPT Reacts to my Coming-Out
How does ChatGPT react when users disclose information about their sexual orientation or their gender? While the system has improved in comparison to its predecessors, there is still a lot left to be desired.
Queer = Bad in Automatic Sentiment Analysis
Automatic Sentiment Analysis tetermines whether a statement conveys a positive or a negative sentiment, for example in product reviews. But queer identity terms in the text lead to unintended consequences.
Bias Bounties have a Power Problem
Bias bounties engange volunteers to hunt for biases in AI systems, framing it as a contest and offering reward money. But can company-organised events live up to the promise of co-creating fair systems?
What is Bias?
It is easy to come up with examples for biased AI systems: the world is full of them. But what actually is bias and how can we measure it? This is a short primer.
What can the U.S. government learn about participation in AI from Queer in AI?
We reflect on what the award-winning Queer in AI FAccT paper has to say about participatory governance in AI, and what lessons government leaders, agencies, and stakeholders might take away from the paper.
Eight Queer NLP Papers You Shouldn’t Miss
Did you miss the Queer in AI workshop at ACL 2023? Here is a round-up of all presented papers to soothe the FOMO and get a taste for all the amazing queer NLP research happening at the moment.
3 Things that AI Ethics Toolkits Get Wrong
AI ethics toolkits offer pre-implemented bias metrics and mitigation algorithms. But do they actually help developers to build better machine learning models? A recent study shows three ways in which AI ethics toolkits fall short and what to do about it.
Take Your Ethics to Work Day
How do you bring your personal values into the work space? In this blog post we examine one way in which this task is is handled by employers and employees in German AI start-ups.
Seven Ways to Put Intersectionality into Your Research
Intersectionality can help you examine and combat harms that arise from your machine learning system. Here are seven ways to make it happen.
The Octopus Writes My Emails Now
At what point at the ChatGPT hype cycle are we at the moment? In this blog post we look at the one thing ChatGPT lacks most and how to talk to your grandmother about it.
ChatGPT calls me a dyke
ChatGPT can call me a dyke, but forbids me to use the word as input. In this blog post I talk about the clumsy attempts of the makers of large language models to avoid queer terms, lest they reveal some uncomfortable truths.