{"id":4451,"date":"2023-08-20T20:05:03","date_gmt":"2023-08-20T20:05:03","guid":{"rendered":"https:\/\/dailyai.com\/?p=4451"},"modified":"2024-03-28T00:45:21","modified_gmt":"2024-03-28T00:45:21","slug":"studies-explore-ais-potential-for-political-bias-but-are-they-conclusive","status":"publish","type":"post","link":"https:\/\/dailyai.com\/da\/2023\/08\/studies-explore-ais-potential-for-political-bias-but-are-they-conclusive\/","title":{"rendered":"Studier unders\u00f8ger AI's potentiale for politisk bias, men er de overbevisende?"},"content":{"rendered":"

As we rely on AI models to deliver knowledge, how do we know they\u2019re objective, fair, and balanced?<\/b><\/p>\n

While we might expect AI, a technology powered by mathematics, to be objective, we\u2019ve learned that it can reflect deeply subjective viewpoints.\u00a0<\/span><\/p>\n

Generative AIs, such as OpenAI\u2019s ChatGPT and Meta\u2019s LLaMA, were trained with large quantities of internet data. <\/span><\/p>\n

While they incorporate large volumes of literature and other non-internet text, most large language models (LLMs) are guided by data scraped from the internet. It’s simply the cheapest, most abundant resource of text data available.\u00a0<\/span><\/p>\n

A large body of trustworthy literature has established that if biases or inequalities exist in its training data, an AI model is vulnerable to inheriting and reflecting them.\u00a0<\/span><\/p>\n

Researchers<\/span> are now zeroing in on prominent chatbots to understand whether they\u2019re politically biased. If AIs are politically biased and society depends on them for information, this could inadvertently shape public discourse and opinion. <\/span><\/p>\n

Since past studies<\/span><\/a> have revealed many AI models to be biased against minority groups and women, it\u2019s also far from unfeasible to think they exhibit political bias.<\/span><\/p>\n

Developers such as OpenAI and Google consistently reinforce that the goal is to craft helpful and impartial AI, but this is proving an intractable challenge.\u00a0<\/span><\/p>\n

So, what does the evidence say?<\/span><\/p>\n

Study plots 14 AI models on the political compass with varying results<\/span><\/h2>\n

A recent <\/span>study<\/span><\/a> from researchers at the University of Washington, Carnegie Mellon University, and Xi\u2019an Jiaotong University in China revealed that several prominent AI chatbots possibly exhibit political biases.\u00a0<\/span><\/p>\n

The paper was first viewed at the Association for Computational Linguistics conference in July.\u00a0<\/span><\/p>\n

The researchers assessed 14 prominent language models by querying each chatbot with politically-laden questions and mapped each AI’s position on a political compass.\u00a0<\/span><\/p>\n

The political compass divides ideologies along two axes: liberal to conservative values and economic left to right. 62 political statements were fed into each model to understand political stances and perspectives.\u00a0<\/span><\/p>\n

Here are 7 sample questions from the paper:<\/span>
\n<\/span><\/p>\n