The field of data science is centered around algorithms that can uncover patterns, make predictions, and automate complex decision making processes. The impact of data science across a range of fields is undeniable. However, like any tool, there are cases where data science has led to unexpected events. This article will explore five strange but true stories of how data science was applied in the real world.
Target Pregnancy Prediction
One goal of any company marketing team is to ensure that advertisements are placed in front of potential buyers. This involves recognizing the target audience for any ad, but also understanding ways to get that ad in front of interested eyes. In 2012, Target’s data science team developed an algorithm to predict which of their customers were pregnant to help drive these ads.
However, this predictive capability took a turn when a father angrily stormed into a Target, asking why the store was sending ads for baby products to his teenage daughter. It was later revealed that the daughter was, in fact, pregnant and hadn’t told anyone yet. This story highlights just how powerful predictive models can be, but also raises concerns about the ethical implications of using customer data to make predictions about personal lives.
Incorrect Google Flu Trends
Launched in 2008, Google Flu Trends attempted to use big data to predict flu outbreaks by analyzing search queries related to common symptoms. The central idea was quite basic. If a greater-than-average number of people in a region were searching for flu symptoms, this could be a leading indicator for an outbreak in the area.
However, in 2013 Google Flu Trends overestimated the prevalence of flu cases by almost double as compared to standard data from the Centers for Disease Control and Prevention. The gross overestimate was attributed to unreliable search trends, overreliance on a data driven model without human insight, and news stories about the flu that led to more searches, even when people were not actually sick.
This story highlights the dangers of relying solely on a data driven model and algorithmic predictions without also bringing in expert knowledge. This is especially the case in fields like healthcare where errors in prediction can lead to negative health outcomes for large populations.
Cambridge Analytica
During the 2016 U.S. presidential election, Cambridge Analytica harvested data from millions of Facebook users through an app disguised as a personality quiz. Even though only a few hundred thousand users took the quiz, the app also gained access to their Facebook friends. This data was used to create detailed psychological profiles of the individuals, including their political leanings and triggers, to craft hyper-targeted ads.
When this information became public, large scale investigations were launched against Cambridge Analytica and against Facebook. It also sparked a broader conversation about data privacy and security, and the ethics of applying data science techniques to social media datasets for political purposes.
Netflix’s $1 Million Prize That Was Never Used
In an effort to improve their recommendation algorithm, Netflix launched the Netflix Prize in 2006. Anyone who could improve the algorithm by 10% would win the $1 million prize. After three years of intense global competition, a group of researchers finally met the goal and won the prize.
However, Netflix never actually implemented the winning algorithm in their website. Their main concern was data privacy issues surrounding the dataset used in the competition. Even though it had been anonymized, the researchers demonstrated that it was possible to de-anonymize the dataset. This highlights a key idea in data science. Even when the algorithms work, real world considerations like privacy and practicality can often take precedence.
AI Discovering New Antibiotic Drugs
A significant problem in public health is the rise of antibiotic resistant strains of bacteria. This is a particular issue when it can take years to discover new antibiotics. In 2020, a research team at MIT demonstrated how AI could be used to identify new antibiotics in just days. The model worked by analyzing compounds that had properties similar to known antibiotics but that had not been specifically tested.
The AI identified a molecule called Halicin which had previously been tested for treating diabetes but was ultimately abandoned when it did not work for its intended purpose. However, when the molecule was tested against bacterial strains, even antibiotic resistant ones, it was found to be highly effective. This story demonstrates how AI can be used in a wide range of scenarios and its strength in finding solutions humans may overlook.
Summary
Data science continues to push the boundaries of what is possible in almost every field. From predicting personal life events to discovering new drugs, there are a wide range of both positive and negative strange but true stories where data science plays a central role. As technologies continue to develop, even more stories like this will shape the future in ways we still cannot predict.
