"Algorithms can affect how millions of people feel, and those people won't know that it's happening" Cathy O'Neil

Narrator: M foreign.

Host: Algorithms can affect how millions of people feel, and those people won't know that it's happening. So wrote Cathy O' Neill in her 2016 book, Weapons of Math Destruction. Welcome to the AI Future podcast, where we hope to add to your understanding of artificial intelligence without getting too techy. Algorithms have become an integral part of our daily lives, influencing everything from social media feeds to search engine results. These sets of computer instructions are designed to perform specific tasks or solve problems, often without human intervention. However, one crucial aspect that has gained increasing attention is their ability to affect the collective emotions of millions of people. Algorithms can subtly shape how individuals perceive the world around them and influence their emotional states, a phenomenon that may remain largely unnoticed by most users. When we scroll through a feed, click like, or simply stare at a headline, our emotional rhythm may change. Yet most of us do not realize the invisible technology at work behind this potential shift. Algorithms have grown increasingly sophisticated in their ability to shape how millions feel. Many remain unaware of the subtle tug they exert on their hearts and minds. In an era where AI has taken center stage, this phenomenon is not only more pervasive, it is also far more powerful and potentially perilous than ever before. Let's start with what we mean by an algorithm. There is no one definition and it will often depend on whom you ask. However, here we can simply think of a basic computer algorithm as a recipe written for a computer to achieve an objective. The computer has been pre programmed to know each recipe. Every step is clear. Like ingredients, each data input is planned for and like a recipe, the resulting output will change based on the data inputs. Therefore, the algorithm appears to be making decisions based purely on data inputs. Artificial intelligence is built on the same kind of step by step logic, only with far more layers and much greater nuance. With AI algorithms, the key difference lies in their adaptability. Unlike basic algorithms, each data input is not necessarily planned for. Our interactions with the technology become another data input that can determine the results. AI algorithms can often understand natural human language. Chatbots or AI agents are increasingly the interface which translate our language into inputs the algorithm can work with. The algorithm adapts to the inputs by comparing it with previous patterns and strategies in its training and our interaction data. The amount of training data can often run into the billions. It then delivers results by predicting what may be the best answer or output for the given inputs it has received. This is the basis of generative AI. It appears intelligent because it can create highly tailored outputs. In theory, algorithms, because they are built using just computer code and data, should be neutral tools. They therefore might be expected to treat everyone the same in reality. The algorithms have often been programmed to tap into deep human psychological principles that may govern individual human perception and emotion. The algorithms can adapt to treat everyone differently. One such psychological principle is cognitive load theory. Algorithms can effectively manage the amount of information presented to users, either increasing or decreasing cognitive load depending on the desired outcome. For example, online news outlets may use algorithms to prioritize sensational headlines that capture attention quickly but might not provide comprehensive context. This can lead to heightened emotional responses such as anxiety or excitement, often based on incomplete information. Moreover, the rise of clickbait culture, which are headlines engineered for high click through rates, can intensify the emotional impact on readers. Such headlines often exploit fear or curiosity to compel engagement, but may fail to deliver substantive information. This practice not only distorts emotional responses, but may erode public trust in journalism over time. Another another psychological principle used is selective exposure. Social media platforms often curate feeds based on past behavior, presenting users with posts that align closely with their previous and current interests. The result? A feedback loop in which individuals are consistently exposed to information that reinforces pre existing emotions. This phenomenon is known as emotional confirmation bias. Over time, this can lock a user into an echo chamber where negative news can amplify anxiety or positive updates can improve optimism, all without the user consciously recognizing what is happening. An example of this may be online news platforms where personalization algorithms may tailor content to each reader's political leanings and interests.

Co-host: While this can increase engagement, it also

Host: may deepen ideological silos. Users may encounter only viewpoints that echo their own, reinforcing emotional attachment to familiar narratives.

Co-host: The result is a fragmented media landscape

Host: where collective understanding fractures along emotional lines. Algorithms therefore seem to be arguably moving

Co-host: from just making decisions to potentially making judgments about us.

Host: AI systems can analyze vast troves of

Co-host: user information, user uh, profiles, browsing history, social media activity, and likes to predict emotional states with remarkable accuracy. Social media platforms epitomize algorithmic influence on emotion. At scale, algorithms may evaluate likes, comments, shares, and time spent to find other content that a user will continue to engage with. Engagement is often psychologically driven by emotional responses to content. Users are more likely to react strongly to stories that elicit joy, anger, uh, or fear. The platform may therefore prioritize feeding further emotionally charged material. Other algorithms might evaluate tags, mentions and engagement rates and promote further content that resonate with users past interactions during heated political debates, this can amplify divisive rhetoric. A single viral tweet may ignite widespread emotional responses across millions of accounts in mere minutes. Emotional contagion is another significant factor. Influenced by algorithms, social platforms often employ features that encourage sharing and interaction, amplifying the emotional tone of content through peer influence. Political propaganda has perhaps seen the most consequential application of AI driven emotional targeting. Political consulting firm Cambridge Analytica in 2018 was famously exposed as uh, sourcing data from Facebook to identify voters and predict psychological profiles and deliver micro targeted ads designed to stir specific emotions, be that fear, hope or resentment, in order to try to shape electoral outcomes. This case exemplifies how AI can turn individual emotional susceptibilities into a weapon for mass persuasion. Another potential aspect of algorithm design that has come under increased scrutiny is whether

Host: some algorithms have been designed to be addictive.

Co-host: As Jonathan Haidt wrote in his 2024

Host: book the Anxious Generation, some platforms are said to be so addictive their algorithms are able to quickly detect whatever it is that makes users pause as they scroll. The algorithms can then provide more of the same. As Adam Mosseri, boss of Instagram, recently testified in a California court, his view was that 16 hours of daily use represented only problematic behavior. Whether such behavior could be considered addictive is something you may want to make up your own mind about. As AI continues to evolve, delivering deeper personalization, real time emotional state detection, and increasingly sophisticated content curation, the stakes for human emotional well being rise accordingly. While algorithms can enrich our lives by providing relevant information and support, they also hold the power to shape collective mood in ways that may be detrimental or manipulative. AI has significantly amplified the potential for algorithms to influence human emotions, primarily through its ability to analyze vastly increased amounts of data and predict individual behavior with remarkable accuracy. AI powered systems can leverage machine learning techniques to identify emotional triggers and tailor content delivery in ways that are both subtle and powerful. This level of sophistication marks a new frontier in emotional manipulation, one that poses significant ethical and societal challenges. Algorithms may therefore have a profound impact on users emotions. So at this point you might ask yourself, do algorithms have too much power? And the answer is probably yes. However, the genie is out of the bottle and it is not going back in. Therefore, understanding how algorithms can influence emotions is crucial for several reasons. First, it helps us recognize the subtle ways technology can shape our thoughts and feelings. Second, it allows us to consider the ethical implications of deploying such systems without adequate transparency or oversight. Lastly, it can provide a foundation for developing more responsible and human centered algorithmic practices. Such insight is crucial for developing strategies to mitigate potential harms and ensure that algorithmic platforms are used responsibly. One can imagine that many of us may have different levels of trust for different professions. For example, there may be a difference

Co-host: between how much we trust a doctor versus how much we might trust a used car salesperson. Perhaps it is time to apply a similar approach to the information that is delivered to us via, uh, algorithms. Some algorithmically driven content may not cause us harm. Other content might. Users should be aware that different types of algorithmically driven content can have different effects on us. Effectively, AI has taken algorithmically driven content from a blunt knife to a sharp scalpel. AI now facilitates hyper personalized content delivery based on an individual's interactions with each associated platform. So by considering the mechanisms behind the algorithms, you may better appreciate the subtle yet significant ways in which technology can shape our emotional experiences and feelings.

Host: Thank you for listening, and we hope you will find interesting other episodes of the AI Future podcast.

Narrator: Sam m.