I know what AI is; that’s ChatGPT!
Do you know what AI means?
“Yes, that’s ChatGPT, right?”
Yes, ChatGPT or other AI tools, but in the broader sense, AI can perceive, process, interpret, and react to data to achieve a specific goal. Moreover, AI can adapt its behaviour by analysing the consequences of its previous actions. In other words, AI can draw conclusions based on learned data, take actions accordingly, analyse the outcomes, adjust its behaviour, and reason and plan all of this.
But what are the risks?
“Okay, isn’t that great? I don’t see the risks.”
Nowadays, it is nearly impossible to utilise online platforms, products, and services without digital surveillance, such as monitoring social media behaviour and tracking locations. As AI can observe, process, interpret, respond to, and reason about these datasets, this combination can lead to the creation of profiles, the monitoring of actions, and the making of predictions. Furthermore, we often voluntarily share information about ourselves or others with AI platforms, such as ChatGPT and Copilot. The companies behind AI are typically not transparent about what information is collected, how it is used, and how it can be deleted. Consequently, our profiles may include personal data related to medical, financial, or political aspects. With this information, the AI company itself, or parties to whom this information is resold, could impose restrictions on our freedom, manipulate us, or even commit identity theft and fraud without our realisation.
“Oh well, that’s my problem. As long as I get a 5.5, haha.”
Oh, really? Did you know that, on average, the male brain is larger and contains more brain cells than the female brain? Suppose someone with a profile indicating they are against women’s rights asks an AI platform a question for which this is the answer. If this platform is influenced by a party with similar beliefs, it could deliberately exclude further context to manipulate this individual and reinforce these ideas. This person might come to believe that men are smarter than women and, therefore, that women are unsuitable for work.
And even if you don’t care, I’m sure that the person you interviewed for your assignment, whose transcript you just entered into AI, does care. Nor will your internship organisation be pleased to find out you included confidential information in AI.
AI as a source, that’s okay, right?
“Okay, I have to think carefully about what information I share with AI, that’s clear. But I usually use AI when I can’t find information on Google, or sometimes when I don’t feel like searching myself. That doesn’t matter, does it?”
AI can analyse and process vast amounts of complex data at high speed to answer your questions, but what it does not do is think ethically.
“Um, what do you mean by thinking ethically?”
Ethics concerns norms and values. Therefore, ethical thinking involves considering whether something is right or wrong, based on these norms and values. This can cover straightforward matters, such as the gravity of child abuse, to more complex situations, like deciding whether statements are justified conclusions or biased. And that is precisely what AI does not do. In fact, many modern AI systems utilise word vectors. To put it simply, you can imagine it as a space where words perceived as equivalent are positioned closer together than those seen as different. For example, consider “man” and “doctor,” and “woman” and “nurse.” Regardless of whether these conclusions are justified or unjustified, AI is incapable of considering that, which can result in social biases in its outputs.
What do you mean, AI isn’t always representative?
“Okay, I have to think carefully about what I share with AI and whether there are any biases in its responses. Is that it?”
Almost! It is also crucial to understand that AI learns from available information, information provided by the company behind the AI or information offered by its users. However, it does not represent all groups of people. In fact, research has demonstrated that women currently use AI less than men. This can lead to inaccurately generated information in certain situations. Furthermore, most AI platforms are designed to provide answers. This implies that if they cannot find an answer in their database or on the Internet, there is a risk they might create their own responses. Therefore, it is essential to fact-check the information obtained from AI at all times.
Wouldn’t inaccurate AI responses be obvious?
“Okay, but if you think logically, you would notice when something is wrong or biased, right? I mean, if we’re discussing the women’s rights example, there are many women who have the same position or a higher position than men. It should make you realise that something is missing or incorrect with the answer, right?”
Figuring that out is called critical thinking, and it doesn’t always happen. AI answers can often appear very convincing and scientifically credible. This leads us to perceive these responses as objective, which results in automation bias.
“And that is?”
That means users think less critically about the results AI provides, compared to how critically they consider answers from humans. You might argue that this is normal because AI is far more objective than humans, but as I previously explained, this is not the case. AI is created by humans and learns from humans. Therefore, our biases are also present in AI. And what if these biases or misinformation are purposefully spread to manipulate us, and reinforce your own biases as well? Would you even realise it?
“No.”
But, how can I use AI ethically then?
“Okay, I understand your point. But, how can I use AI ethically?”
I know, using ethically can seem rather vague, but it is pretty straightforward to implement. Here are a few steps to follow.
“Hm, alright, these points seem pretty logical. I can work with that.”
Using AI ethically, I can do that!
AI is fast, easy, and has many possibilities. Therefore, we certainly don’t want to discourage its use; however, it is essential to be aware of its drawbacks. For instance, neglecting to consider what you share with AI may result in your personal data being used for profiling, monitoring, and making predictions. If this information falls into the wrong hands, it could be used to restrict your freedom, manipulate you, or commit identity theft or fraud. Furthermore, not all AI responses are accurate or representative of everyone, and they may reflect societal biases due to word vectors. Thus, it is crucial to think critically and ethically when using AI. Consider carefully what you share, how you interpret the answers, and always conduct further research!
“Okay, thnx for these tips! I certainly wasn’t planning to stop using AI either, but I will be more conscious from now on!”
Anyl Snelder