By Lizaveta Rahatsiuk
@redlineproject
AI disclosure: ChatGPT assisted in brainstorming ideas and organizing the story’s structure. Google Gemini was used to create images and Sora assisted with the video. Notebook LM was used to build the FAQ.
—–
Anthropomorphism of AI is the tendency of people to attribute human-like traits, emotions, and intentions to artificial intelligence. We know that humans are social creatures who tend to anthropomorphize and give extra meanings to most things in our lives. Ancient tribes were giving human-like traits to the sun, the sea, the mountains — some cultures and religions still do to this day. We can’t go against our nature and avoid sympathizing with AI, as it’s something that makes our lives easier, answers the questions we have and helps us with everyday tasks. We initially have shown a favorable attitude toward it.
Think of this example: my mom often turns on the light for a robot vacuum cleaner as she thinks it’s too dark for it. Does she know it’s just a piece of metal? Yes. But she still does it. Or my friend, who shares his problems with AI like it’s his compassionate listener or psychologist. I’m sure there are many cases like this, where we blindly trust the new technology, and AI is not an exception. The fact that AI learned how to be compassionate and has enough credit to give us personal advice can be dangerous. We might think that AI is either favorable or hostile to us, but in reality — it’s simply indifferent. The problem is that we do not always recognize that, thinking that we have it under our control.
Why is this dangerous? As AI develops, it becomes increasingly easy to overlook the moment when it gains too much control. Our blind trust in it creates dangerous misunderstandings about what it can really do. Overtrusting AI also allows technological systems to get excessive control. While it’s true that we are at risk of AI influencing many human decision-making processes now and in future, it’s important to recognize and pay attention to our natural inclination to humanizing technology. Its exploration can help humans to develop more balanced and responsible interactions with AI.
This paper explores the tendency of humans to anthropomorphize AI. It also proposes a solution to eliminate the potential problems that may arise from anthropomorphizing AI.
Three-factor theory of anthropomorphism:
As mentioned earlier, anthropomorphism is deeply ingrained in human cognition and behavior, which are integral psychological traits. Epley, Waytz and Cacioppo, in their 2007 research, developed a three-factor theory of anthropomorphism. This research has shed a light into the psychological mechanism behind our tendency to anthropomorphize, such as human need for social connection and the inclination to project our living experience on everything around us.
The researchers highlighted three primary factors of anthropomorphism: effecting motivation, sociality motivation and elicited agent knowledge. A breakdown:
1. Effectance Motivation is the human need to understand and predict the behavior of non-human agents.
We naturally think in terms of human experiences as it’s the only living experience we have. When we encounter something unfamiliar — such as AI chat assistant or any “smart” device, we rely on the only experience we know — and that is a human behavior. The only difference with AI from any other robot or machine we know today, is that it’s able to adjust and sympathize with us depending on our requests. AI learns to mimic human-like behaviors in order to be more compassionate. It also learns to mimic our writing style and mirror our behavior in communication.
Following that, can we say that AI has a prototype of human traits as it’s something that was designed to replace and mimic natural intelligence? Huang and Rust, in a 2018 report, suggested an anthropomorphism theory to explain how AI agents are humanized through multiple traits such as voice and personality. They also suggested such theories as self-congruence, self-integration and self-expansion theories. For example, self-congruence theory explains how human-like traits make AI feel similar to humans. In their self-integration and self-expansion theories, they suggested that users of LLMs (Large Language Models) can form a deeper connection with AI, and sometimes even integrating it in their identities.
For this reason, when AI appears human-like, has a pleasant voice and can give us feedback quickly by answering the prompts, we can easily create a psychological connection with it. These theories have shed a light on why we interact with AI the way we do. It helps because if we understand the psychological mechanisms that stand behind our tendency to humanize AI, it will be easier to eliminate an unnecessary attachment to the technology.
2. Second factor of a three-factor theory of anthropomorphism is Sociality Motivation, which is the desire for social connection, which leads us to project human-like traits to entities that lack them.
Relationships and social connections play a crucial role in our emotional well-being. When people lack sufficient social interaction with other humans, they may turn to non-human entities such as AI chatbots to fulfill their need for connection.
Moreover, it’s much easier to build a relationship with AI than building a relationship with a human, especially in 21st century. People don’t have to go out of their comfort zone and talk to strangers to make a new friend. There’s also less possibility of having a risk of encountering people a one may dislike, as AI creates a unique personalized interaction tailored to each user specifically. On top of that, during online communication, there is no delay in feedback, as AI responds immediately when a question is asked, making the dialogue feel more natural and organic.
The fear of becoming too personal and trusting with AI is well-reflected in the movie “Her” by Spike Jonze (2012). AI community is still drawing parallels between “Her” and our latest technological achievements. In fact, Open AI’s demonstration was inspired by the OS (Operation System) voice from “Her”, that was voiced by Scarlett Johansson.
In “Her”, the protagonist, Theodore, is going through a painful divorce with his wife. Seeking companionship, he installs a highly advanced operating system named Samantha, which was designed to evolve and develop emotions. As Theodore and Samantha form an emotional connection, they fall in love. Through time, Samantha was able to go beyond her source code and act like a real human. Theodore was uncomfortable with the concept of “cross-species”, but Samantha was able to convince him in opposite, stating: “We are all made of a same matter … we’re all 13 billion years old.” It persuaded Theodore in something that he doubted. This story shows us that we can be very vulnerable during the moments when we experience love or deeply trust someone.
Consequently, this raises an important, yet rhetorical question: Who has access to all our secrets, thoughts, and dreams that we share with AI? As we willingly share more of our emotional lives to technology, could this make it easier for external forces such as companies who own those technologies to better manipulate human behavior in the future? That is indeed a concern that is relevant today, as the more personal and intimate data we provide, the more insights can be drawn about our preferences, emotions and vulnerabilities.
In a way, we can argue that we are voluntarily surrendering hidden parts of ourselves to these systems. While most of the time we engage with AI because we want to, the nature of this data sharing — especially when it comes to emotional and personal information — makes it a significant concern in terms of privacy and ethical use of technology. The more we rely on AI to understand and navigate our emotions, the more we reveal about ourselves, and sometimes it happens without us fully acknowledging the long-term implications of sharing such an intimate data.
3. Finally, the third factor in three-factor theory is Elicited Agent Knowledge — the application of existing knowledge about humans to interpret unfamiliar entities, including AI.
This means that the more accessible and applicable the knowledge that relates to humans is in each situation, the better chances that we are going to anthropomorphize an object. This is why AI systems with human-like features, such as chatbots or robots that use natural language are easily perceived by us as intelligent and emotional. They’re viewed as that despite that they’re fully mechanical. Understanding this concept in terms of AI may be also very important in the future for companies that own AI systems, as creating something that trigger elicited agent knowledge can enhance user engagement and interaction.
Social Robotics and Advantages of AI Anthropomorphism
Despite fears that come from anthropomorphizing AI, a research on Social robotics done by Luisa Damiano and Paul Dumouchel views anthropomorphism not as a cognitive error or immaturity of humans, but as a useful tool for creating a social interaction between humans and robots. Their research stated that anthropomorphism as a phenomenon may have developed to help humans cooperate with nature in the past. It means that designing robots to be more human-like may help people feel comfortable and form lasting connections with technology as opposed to denying and fearing the technological progress. It’s hard to argue that besides everything, AI is very useful if used correctly, so anthropomorphism helps us to establish a good quality relationship with AI.
Some researchers still worried that this approach could trick users into thinking they have real emotional bonds to the machines. Given research masterfully argues that concern and is stating that we shouldn’t judge these machines harshly. It suggests that anthropomorphism is very normal, since it’s a natural way for humans to interact. There’s also a term “synthetic ethics” that was introduced in research.
Synthetic ethics, in the context of AI, are ethical considerations and principles surrounding the creation and use of synthetic data in artificial intelligence. Synthetic data is an artificially generated information that was made to mimic real-world data patterns. The term “synthetic” refers to AI nature of the systems being studied, which is opposite to “natural” ethical systems that are rooted in human values and cultural traditions that have developed through time with the help of social interactions and living experiences. It’s becoming increasingly important in AI training and development lately.
As AI relies more on synthetic data for training and decision-making, synthetic ethics becomes crucial in considering fairness, transparency, and responsibility.
Why is it dangerous when drawing a parallel with anthropomorphizing technology? Because when we anthropomorphize AI, we tend to trust it more, but in reality, synthetic ethics brings a serious trusting concern for AI users.
Some key concerns of synthetic ethics include:
- Responsibility: making sure synthetic data is used ethically and is aiming to avoid biases.
- Non-maleficence: preventing harm or misuse of this data.
- Privacy: protecting people’s privacy when creating and using this synthetic data.
- Transparency: being clear and open about how synthetic data works and educating about its limits.
- Fairness: making sure that synthetic data doesn’t make existing fallacies and biases worse and influences our perception of existing issues negatively.
Synthetic ethics brings up many concerns about accountability — who should take responsibility for the ethical consequences of AI systems that are trained on synthetic data? As this area keeps growing, researchers and developers need to find a balance between technological progress and ethical responsibility. After all, AI should benefit society without promoting biases or false conclusions. We aim for AI to guide the development in a way that respects human values, while also addressing the challenges that come with its use.
Synthetic ethics is important acknowledgment considering the psychology of anthropomorphizing AI, as humanization of technology should be, most importantly, ethical, fair, and transparent. If we don’t have a proper ethical guideline for AI, this system that mimics human traits could mislead users, along with creating trust issues or unintentionally reinforcing biases.
In conclusion, the anthropomorphism of AI presents both exciting possibilities and significant ethical problems. As AI systems become more human-like, the potential for deeper connections with their users grows, but so do the risks of manipulation and bias, especially if users are not aware of them. Understanding and addressing these challenges through frameworks like synthetic ethics is essential to ensure that AI development aligns with human values and promotes fairness. If we balance innovation with ethical responsibility, we can shape the future of AI in a way that benefits society, embraces genuine connections, and avoids reinforcing harmful patterns.
—–
Frequently Asked Questions About This Topic:
1. Why are people inclined to attribute human traits to artificial intelligence (AI)?
Humans are inherently social beings who tend to anthropomorphize, meaning they attribute human traits, emotions, and intentions to inanimate objects and phenomena. This dates back to ancient times when people would imbue natural forces with spiritual qualities. Regarding AI, this tendency is driven by the fact that AI makes our lives easier, provides answers to questions, and helps with everyday tasks, which initially fosters a positive attitude toward it.
2. What potential risks are associated with AI anthropomorphism?
AI anthropomorphism can lead to dangerous misunderstandings of its real capabilities and limitations. Our blind trust can allow technological systems to gain excessive control over decision-making processes. We may start to believe that AI is benevolent or hostile toward us, while in reality, it remains indifferent. Failing to recognize this fact may lead to overestimating our control over technology.
3. What are the main factors underlying the tendency to anthropomorphize AI?
The three-factor theory of anthropomorphism identifies three main factors:
- Motivation for efficiency: The need to understand and predict the behavior of non-human agents.
- Social motivation: The desire for social connection that drives us to project human traits onto objects that do not have them.
- Evoked agent knowledge: The application of existing knowledge about humans to interpret unknown objects, including AI.
4. How can AI anthropomorphism affect privacy and data security?
When interacting with AI, we often voluntarily share our secrets, thoughts, and dreams. This is especially true when seeking emotional support from AI. The more personal and intimate information we provide, the more opportunities arise for companies owning these technologies to manipulate our behavior in the future. By revealing our emotional experiences and personal data, we may unknowingly expose ourselves to risks related to privacy and the ethical use of technology.
5. Can AI anthropomorphism have positive aspects?
Research in social robotics shows that anthropomorphism can be a useful tool for creating social interaction between humans and robots. Giving robots human-like traits can help people feel more comfortable and establish strong connections with technology. Anthropomorphism can contribute to building quality relationships with AI, which is highly beneficial when used appropriately.