Social media defines you. What if they are completely wrong?
Pinterest and Threads seem to think I'm 60 years old or going through menopause. If you spend time in those strange content worlds, you can end up believing anything.
The experience of receiving personalized recommendations on applications like Pinterest raises questions about the effectiveness and relevance of these systems in the era of artificial intelligence. A few years ago, a user shared their frustration at canceling their wedding and still receiving hairstyle and decoration suggestions that no longer interested them. Although the responsibility for this phenomenon does not solely lie with Pinterest, the feeling that social media offers outdated content has become common.
In a context where tech companies use advanced AI systems to enhance their advertising performance, one might think that recommendations would have also evolved. However, that is not always the case. Despite the sophistication of the algorithms, many seem unable to shed old data and offer recommendations that do not match the user’s current reality. Three years after the "miscarriage problem" that Pinterest acknowledged, the author still receives inappropriate suggestions, such as hairstyles for older people, when in fact they are a millennial.
The impact is not limited to Pinterest; other platforms like Meta's Threads also face similar challenges. Threads, designed to be a distinct application, displays a site of mostly text-based updates and promotes interactions that have led the author to receive content related to menopause, despite it not being applicable to their situation. Even more puzzling is that this content seems to be filtered based on past interactions, but often inaccurately.
Meta explained that the content a user receives is based on various signals, such as accounts and posts they have previously interacted with. This highlights the lack of accuracy that can accompany the use of old data. Considering that the user has not logged into Pinterest for over a year, the application can only provide erroneous and outdated content.
The conclusion is that recommendation platforms require active user interaction to improve their accuracy, placing the burden on the user to update their preferences. This reliance on users to tailor the recommendation algorithms to their current situation signifies a shift in how we navigate our online identities. Instead of taking control of our digital experiences, we are forced to follow the recommendations of algorithms that, at times, do not reflect who we truly are.
In summary, although the goal of these technologies is to facilitate access to relevant content, current realities show that obsolescence and a lack of genuine communication with users persist, highlighting a significant challenge in the evolution of digital platforms.
