Introduction to Machine Learning
What in the world is AI and ML?
Artificial intelligence (AI) and machine learning (ML) are sometimes silent, and often invisible systems that are omnipresent in our daily lives. If you are reading this, it is likely that ML contributed to your arrival at this exhibit, be it through the results you received on a search engine, or through a link in an email that was not marked as spam. Onuoha and Nucera (2018) define artificial intelligence as “the theory and development of computer systems able to perform tasks that normally require human intelligence” (p. 8) and machine learning as “a branch of artificial intelligence in which a computer generates rules and predictions based on raw data that has been fed into it” (p. 9). AI is used as a tool to automate tasks that humans previously performed, to complete a job quicker or more efficiently (Onuoha and Nucera, 2018). ML algorithms are a subset of AI and are used to automate tasks beyond what humans are able to perform by finding patterns within information (Onuoha and Nucera, 2018). Once a machine can find the pattern, it can then predict what follows (Onuoha and Nucera, 2018). In a world of information overload, ML is helpful to dig through large amounts of data to provide customized recommendations based on things we have liked in the past.
Through this exhibit we will demonstrate how AI and specifically ML is woven into today’s cinematic, literary and music cultures. The most obvious way that users interact with ML is through applications that sell, provide or promote pop culture. In this exhibit we have performed case studies on three such applications: Netflix, Goodreads, and Spotify. All three platforms use ML to generate recommendations for new movies, shows, music and books to consume, in order to keep the user engaged, satisfied and returning to the application.
Recommender systems
Recommender systems utilize ML to help users make choices by providing a reduced list of options (i.e., recommendations) that are algorithmically personalized. Users effectively want to be presented with the best choices, for the least amount of work. Once a recommender “knows” a user, it can give nudges and recommendations at the right time to fulfill tasks, which can greatly and positively influence an individual’s decision-making, but also gives more power to corporations and their business interests (Schrage, 2021). However, users should be cognizant that most recommendations are biased, and should think critically about who recommenders are benefitting.
Recommender systems are the key to success for many of the biggest digital platforms, and a well-functioning system leads to loyal customers and significant business growth (Schrage, 2021). The best recommendation engines are machine learners that become smarter over time, so the more a recommender is used, the better the recommendations will be (Schrage, 2021). The way a user utilizes the platform, such as browsing, rating and purchasing, provides the data that helps the recommender learn.
Recommender systems redefine both insight and influence in the age of algorithms and have the ability to affect human behaviour (Schrage, 2021). A successful recommender is trusted by users, so they will be more likely to try something out of the ordinary (Schrage, 2021), such as picking up a book by an unknown author or watching a new movie genre. Knijnenburg et al. (2012) explain that users spend more time on programs that utilize personalized recommenders. Further, users perceive a system to be more effective when browsing activity is reduced (Knijnenburg et al., 2012). Happy clients that continually return to the platform are good for business, because recommender systems are designed to benefit both the applications and end user.
The three case studies outlined in this exhibit—Netflix, Goodreads, and Spotify—have ML recommenders that are an important component to their business model. Because of their recommenders, these platforms are able to recruit new members, and maintain a large number of clients. Effective recommender systems do not just maintain a client base, but they can also change a user’s behaviour. When recommenders provide a highly personalized experience, satisfied users may end up consuming more content, and as a result possibly viewing more ads. Users might be compelled to watch or listen to one more episode, or follow an interesting link, which in turn leads them to spend more time on the platform. Successful recommendations will result in higher rated content, helping to improve the user’s personalized recommendations and creating loyalty to the platform as will be further discussed in the exhibit.