AI Recommendation Algorithms Can Worsen Loneliness

Cottonbro / Pexels

New research suggests that artificial intelligence (AI)-powered algorithms like personalized searches, recommendation systems, and AI-powered curation can make us more isolated and polarized, or worsen one's mood. While AI can reinforce and amplify a rigid mindset, it is not responsible for the tendency to want to see information that is consistent with our preexisting views. The responsibility is on us humans to identify and take steps to shift from a self-limiting rigid mindset to a growth mindset.

One of the most common applications of AI is algorithms that provide recommended posts on online platforms, including social media. These AI algorithms automatically filter out information and are built to serve users individualized content based on past engagement and preferences. This results in a phenomenon of "information cocoons," "filter bubbles," or "echo chambers" in which people are less likely to encounter content that is different from their own preferences and beliefs.

Users make active choices based on their preferences, and the algorithms respond by serving up similar information. The underlying wish to be surrounded by familiar and similar beliefs is part of a tribal mentality that has existed long before AI. Humans are also in charge of designing these algorithms and online platforms, so it would be unfair to "blame" AI for this bias. It is important to be aware that this bias is amplified by AI through a feedback loop that reinforces the rigidity of our preexisting beliefs.

In the 2006 book Infotopia: How Many Minds Produce Knowledge, legal scholar Cass Sunstein argued that people focus on topics that interest them and effectively create a "personal daily newspaper" that enables them to exclude or ignore other viewpoints and content. These "information cocoons" create insular environments or "echo chambers" in which one's own opinions become stronger and more entrenched. AI algorithms reinforce these personalized echo chambers since frequently these algorithms are built to maximize user engagement rather than information diversity. The psychological impact of these individualized echo chambers can lead to social isolation, polarization, and reinforcement of a rigid mindset.

Information cocoons have been linked to higher levels of loneliness and depression, which can be overcome by family and social support. However, in the absence of active family or social support, people can end up feeling more isolated in the echo chambers of their own beliefs. This effect is particularly worse for those who already experience depression, loneliness, and anxiety.

Information cocoons can worsen depression or anxiety by reinforcing negative schemas or core beliefs. Early life experiences combined with genetic and epigenetic modifications can create these negative automatic beliefs or schemas, and user-generated data will be biased to reflect these schemas. AI algorithms then lead to the proliferation this negative bias, creating information cocoons that constantly provide more "validating evidence" or data that reinforces one's negative beliefs about oneself, others, and the world. The schemas underlying the depression take hold with even greater rigidity and inflexibility with time on online platforms. In other words, downward-spiral negative thinking can be amplified by these AI recommendation algorithms through this human-AI feedback loop.

The human-AI interaction underlying these recommendation algorithms has the potential risk of reinforcing and worsening rigid negative mindsets. This is particularly problematic for individuals who have anxiety or depression. People can overcome this by actively seeking to become conscious of and disrupt these self-limiting beliefs, but this requires awareness and a willingness to change. Leaving the information cocoon is possible by empowering oneself to seek out and interact with new information and content that directly counters negative beliefs and instead promotes self-care, healing, and growth.

Marlynn Wei, MD, PLLC © Copyright 2024