They say holidays are a time of relaxation. Well, they are – but only after struggling the overbearing battle of buying the so called “perfect gift” and then ultimately accepting that it will never happen.
It’s the annual tale. Around 81% of Americans have difficulties have gift shopping during the holiday season. Whether it be a gift for a close friend or a thank you gift for a teacher, that hesitation we have when buying stuff for ourselves multiplies when we begin looking for a gift for other people.
But recent improvements to online shopping aim to make the process quicker, with features such as recommendation engines based on user history.
Recommender systems generate meaningful recommendations to a collection of users for items or products that might interest them by using machine learning techniques to get a holistic picture using other people’s shopping history and product ratings. As explained by Kaushik Pal of Techline, recommender systems work by collecting a large amount of data from users using their shopping history and and preferences to generate ratings and recommendations.
And truth be told, sometimes recommender systems can be wonderful, helping us narrow down the endless pen options to find that one unique pen that we are trying to give to that friend who is a stationary junky and an endless hoarder. That perfect pen is found by the automated system which analyses all the disappointing and satisfying decisions made by previous consumers. Every single purchase.
But what are the costs of online shopping specifically these recommendation engines?
Tell me about yourself.
Privacy – or lack thereof.
Juniors Madi Anderson and Arjun Subramonian express similar opinions regarding privacy when it comes recommender systems. As an avid online shopper, Anderson does not find the privacy of online shopping in general that concerning.“I’m not really worried about privacy,” she said, “but my mom always tells me to be careful online.”
For many students like Anderson, privacy in online shopping is not a major concern as many are not too worried about their shopping history. In fact, a recent survey of 42 MVHS students conducted through Google Forms showed that over 47% of MVHS students are not concerned over privacy of online shopping.
“There are terms and conditions when you shop at places like Amazon and they don’t keep the data forever,” said Subramonian, who is heavily interested in the software behind online shopping systems. Indeed, many of these online sites take into consideration the user’s privacy by erasing data over time or letting the consumer know how their preferences will be recorded.
But even keeping record of purchases temporarily can be concerning.
Sophomore Ashley Lin who agrees that privacy must be kept even on a virtual platform says, “Sometimes, people don’t want all their shopping history up for grabs but they don’t have a choice.” Indeed, the lack of individual options is a driving force behind the privacy concerns of recommender systems.
Anything I can do to help?
Privacy concerns aside, another underlying question remains: do recommender systems really benefit consumers?
Subramonian thinks otherwise. “Bottom line it’s for the companies,” he said.“It always is.” He finds that although a generalization of preferences can help the consumer get a sense of good products to buy, ultimately for sites like Amazon and Netflix, these systems can be skewed to promote only certain products.
Anderson finds that similarly recommender systems can be beneficial in only certain situations, like narrowing down options. “It gives me an idea of what I can buy,” says Anderson, “but I don’t typically use them when buying gifts for other people.” When it comes to recommender systems, she has seen the recommendations are more in tune to her likes rather than those of who she is gifting, which is not necessarily helpful.
MVHS students in general have a similar opinion. Only 10-20% of MVHS students use these recommendations though around 43% of students recognize the benefits of recommender systems.
Why do we not eventually trust that “special”, “extra comfortable” black pen that “many other customers purchased” (other than the cost of course)?
It’s known as the “Peppa Pig Problem.” Like with Netflix that fills your feed with kids shows after your little cousin sister watched too much tv, online recommendations are often too intune to special groups or demographics and cannot give the intended holistic recommendation. Sometimes the special groups might be the online provider itself. For example, sites like Amazon prime promote their original shows more often than other shows, showing a conflict of interest.
Another reason why these recommendations are not trusted or inaccurate is because of the “rating bubbles” they are based upon. The MIT Sloan Management Review finds that social influence and herd behaviour results in most reviews, which are analyzed by machine learning, that are more skewed to a certain extreme sentiment, writing that “It turns out that online ratings tend to be disproportionately positive. The distributions of product ratings on Amazon.com include far more extreme positive (five-star) than negative (one-star or two-star) or generally positive (three-star or four-star) reviews.” In other words, the reviews that online recommender systems are based on are often not the best source to categorize products.
Ultimately, although online recommendations has potential to simplify the decision and make our lives, it still raises privacy concern and accuracy issues. One thing does is clear. The battle for the perfect gift is still on. Everyone, onward!