{"id":15722,"date":"2025-08-20T18:47:08","date_gmt":"2025-08-20T08:47:08","guid":{"rendered":"https:\/\/rationalemagazine.com\/?p=15722"},"modified":"2025-08-20T18:47:08","modified_gmt":"2025-08-20T08:47:08","slug":"its-fantasy-to-think-ai-can-be-free-from-bias-and-ideology","status":"publish","type":"post","link":"https:\/\/rationalemagazine.com\/index.php\/2025\/08\/20\/its-fantasy-to-think-ai-can-be-free-from-bias-and-ideology\/","title":{"rendered":"It’s fantasy to think AI can be free from bias and ideology"},"content":{"rendered":"

In July, the United States government\u00a0made it clear<\/a>\u00a0that artificial intelligence (AI) companies wanting to do business with the White House will need to ensure their AI systems are \u201cobjective and free from top-down ideological bias\u201d.<\/p>\n

In\u00a0an executive order<\/a>\u00a0on \u201cpreventing woke AI in the federal government\u201d, President Donald Trump refers to diversity, equity and inclusion (DEI) initiatives as an example of a biased ideology.<\/p>\n

The apparent contradiction of calling for unbiased AI while also dictating how AI models should discuss DEI shows the notion of ideologically free AI is a fantasy.<\/p>\n

Multiple<\/a>\u00a0studies<\/a>\u00a0have shown most language models skew their responses to left-of-centre viewpoints, such as\u00a0imposing taxes on flights, restricting rent increases, and legalising abortion<\/a>.<\/p>\n

Chinese chatbots such as DeepSeek, Qwen and others\u00a0censor information<\/a>\u00a0on the events of\u00a0Tiananmen Square<\/a>, the\u00a0political status of Taiwan<\/a>, and\u00a0the persecution of Uyghurs<\/a>, aligning with the official position of the Chinese government.<\/p>\n

AI models are not politically neutral nor free from bias. More importantly, it may not even be possible for them to be unbiased. Throughout history, attempts to organise information have shown that one person\u2019s objective truth is another\u2019s ideological bias.<\/p>\n

Humans struggle to organise information about the world without distorting reality.<\/p>\n

Take cartography, for example. We might expect maps to be objective \u2013 after all, they reflect the natural world. But flattening a globe onto a two-dimensional map means having to distort it somehow. American geographer Mark Monmonier\u00a0has famously argued<\/a>\u00a0maps necessarily lie, distort reality, and can be tools for political propaganda.<\/p>\n

Think of the classic world map that uses\u00a0the Mercator projection<\/a>, hung in every primary school classroom. It converts the globe into a cylinder and then lays it flat. I grew up thinking Greenland must be massive compared to the rest of the world.<\/p>\n

In fact,\u00a0Africa is 14 times larger than Greenland<\/a>, despite appearing to be roughly the same size on this type of map.<\/p>\n

In the 1970s,\u00a0German historian Arno Peters argued<\/a>\u00a0Mercator\u2019s distortions\u00a0contributed to a skewed perception<\/a>\u00a0of the inferiority of the global south.<\/p>\n

Such distortions could be an analogy for the current state of AI. As Monmonier wrote in his book\u00a0How To Lie With Maps<\/a><\/em>: “a single map is but one of an indefinitely large number of maps that might be produced for the same situation or from the same data.”<\/p>\n

Similarly, a single large language model\u2019s response is one of an indefinitely large number of responses which might be produced for the same situation or from the same data.<\/p>\n

Think of the many ways a chatbot could formulate a response\u00a0when prompted<\/a>\u00a0about something like diversity, equity and inclusion.<\/p>\n

Other historic attempts at organising information have shown the bias of their designers and users too.<\/p>\n

The widely used Dewey decimal classification (DDC) system for libraries, published in 1876, has been known to be\u00a0racist<\/a>\u00a0and\u00a0homophobic<\/a>.<\/p>\n

Throughout history, attempts to organise information have shown that one person\u2019s objective truth is another\u2019s ideological bias.<\/strong><\/p><\/blockquote>\n

Over the course of the 20th century,\u00a0LGBTQIA+ books have been categorised<\/a>\u00a0under Mental Derangements, Neurological Disorders, or Social Problems in the DDC, with more recent efforts being made to\u00a0eliminate outdated and derogatory terms<\/a>\u00a0from the classification.<\/p>\n

Under Religion, roughly 65 sections out of 100 are dedicated to Christianity<\/a>\u00a0because the library where the classification was originally developed had a strong focus on Christianity. While\u00a0Islam has an estimated 2 billion followers<\/a>\u00a0to Christianity\u2019s 2.3 billion today, in the DDC Islam only has a single section dedicated to it.<\/p>\n

The large language models that power AI chatbots are trained on countless pieces of text, from historical works of literature to online discussion forums. Biases from these texts can unknowingly creep into the model,\u00a0such as negative stereotypes of African Americans from the 1930s<\/a>.<\/p>\n

Just having raw information is not enough. Language models must be trained how to retrieve and present this information in their answers.<\/p>\n

One way to do this is to have them\u00a0learn to copy how humans respond to questions<\/a>. This process does make them more useful, but studies have found it also makes them align\u00a0with the beliefs<\/a>\u00a0of those\u00a0who are training them<\/a>.<\/p>\n

AI chatbots also use system prompts: instructions that tell them how to act. These system prompts are, of course, defined by human developers.<\/p>\n

For example, the system prompts for Grok, the AI chatbot developed by Elon Musk\u2019s company xAI,\u00a0reportedly instruct it to<\/a>\u00a0\u201cassume subjective viewpoints sourced from the media are biased\u201d and to not \u201cshy away from making claims that are politically incorrect, as long as they are well substantiated\u201d.<\/p>\n

Musk launched Grok to\u00a0counter his perceived \u201cliberal bias\u201d<\/a>\u00a0of other products such as ChatGPT. However, the recent fallout when Grok\u00a0started spouting anti-semitic rhetoric<\/a>\u00a0clearly illustrates that attempts to correct for bias necessarily replace it with\u00a0a different kind of bias<\/a>.<\/p>\n

All this goes to show that for all their innovation and wizardry, AI language models suffer from a centuries-old problem. Organising and presenting information is not only an attempt to reflect reality, it is a projection of a worldview.<\/p>\n

For users, understanding whose worldview these models represent is just as important as knowing who draws the lines on a map.<\/p>\n

 <\/p>\n

This article was originally published in <\/strong><\/em>The Conversation<\/strong><\/a>.<\/strong><\/em><\/p>\n

Photo by Aerps.com <\/a><\/strong><\/em>on Unsplash.<\/strong><\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

In July, the United States government\u00a0made it clear\u00a0that artificial intelligence (AI) companies wanting to do business with the White House<\/p>\n","protected":false},"author":838,"featured_media":15729,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"footnotes":""},"categories":[65],"tags":[562],"coauthors":[797],"class_list":["post-15722","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-science-health","tag-artificial-intelligence"],"acf":[],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/posts\/15722","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/users\/838"}],"replies":[{"embeddable":true,"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/comments?post=15722"}],"version-history":[{"count":4,"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/posts\/15722\/revisions"}],"predecessor-version":[{"id":15728,"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/posts\/15722\/revisions\/15728"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/media\/15729"}],"wp:attachment":[{"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/media?parent=15722"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/categories?post=15722"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/tags?post=15722"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/rationalemagazine.com\/index.php\/wp-json\/wp\/v2\/coauthors?post=15722"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}