Desktop version

Home arrow Philosophy

  • Increase font
  • Decrease font


<<   CONTENTS   >>

How big data has been leveraged by industry

Ostensibly, the commercial value of tracking and identifying the different strategies that users employ across varied digital spaces lies in the increased ability of a company to improve the user’s experience, or to better promote certain actions on the part of the user, such as making a purchase or continuing to use the company’s services. The goal of such a robust data collection effort is to build analytic models of each user, or user data profiles. These profiles are largely based on the types of strategies that users employ on websites and other digital spaces (Wang, 2008). Thus, as users continue to interact with digital services, individual user models or profiles are continually refined with the intent of developing databases of robust profiles. Such profiles can be used to identify with what features of a digital service users are engaged, whether users regularly achieve their intended outcomes, whether they have difficulty in completing some or all of their intended actions, or whether the software environment is hard to use or otherwise not intuitively designed based on a user’s intentions (Kardan & Ebra-himi, 2013).

The thought of large companies that own comprehensive dossiers on individual users that account for every single click and movement might indeed seem a little like science fiction. However, website and app providers actually often allow third parties to track their own users across platforms, devices, and websites. Such broad, multi-platform sharing of data between services and websites occurs on virtually every website today, which can readily reveal the types of various browsing, buying, and content viewing strategies of users to companies who desire such information. In addition, the prospect of only being tracked in digital spaces is increasingly less likely given today’s commercial technologies, as many mobile device apps readily relay users’ physical location data back to companies.

A common example of multi-platform tracking is highlighted by some web analytics tracking services, which are available for free to website and app providers. Such services find widespread use on websites or mobile apps across all industries today (Clifton, 2012). In two of the most common examples, Google (via Googles Analytics service) and Facebook (via Facebook Pixel tracking) provide small snippets of programming code to website and app providers, which are placed on websites that are outside of Google or Facebooks ownership or control. As a result of embedding these snippets of tracking code, Google and Facebook provide software administrators with robust analytics tools that the administrators can use to better understand visitor behavior, including the strategies that they use on the website (Clifton, 2012). However, in return for the monetarily free cost of providing the tracking service, this data on the behaviors on websites is aggregated and returned to Google and Facebook. Such data from external websites and services allows these companies to enrich the user profiles with greater resolution, which is a hallmark of Big Data.

This type of multi-site tracking is also not just limited to capturing data on the strategies employed by users on websites. For example, mobile phones often collect and transmit a host of data to third parties that is valuable for inferring the types of behaviors and strategies that users perform as they go about their everyday lives. Given that an application on a mobile device has the appropriate level of permissions on the mobile device, various dimensions of data are shared from mobile devices to app providers as they are collected from the devices sensors and interaction with nearby internet-connected devices. These multiple data dimensions include a user’s location, a user’s physiological attributes (e.g., measuring steps taken, heart rate, and rate of movement as tracked by a device’s accelerometer, compass, and gyroscope), and environmental factors (e.g., temperature and ambient sound). As a result, the data that composes a person’s digital fingerprint is no longer exclusively digital: physical and digital lives are increasingly bridged in the era of Big Data. As a result, seemingly more daily activities can be observed and analyzed in Big Data contexts.

By collecting comprehensive, cross-platform data on users’ tastes in content, the entertainment media industry has embraced Big Data approaches to suggest new content to users. Primarily, home entertainment video and TV streaming media services such as Netflix, YouTube, Hulu, and Amazon Prime have embraced Big Data efforts, such as collection and analysis of user viewing preferences by engaging in large-scale data collection and user modeling efforts (Bennett & Lanning, 2007; Gomez-Uribe & Hunt, 2016). However, other media providers have also used Big Data to make content suggestions to individual users based on their prior viewing patterns and inferences on their interests, such as social media content like Twitter, Facebook, and Snapchat (Gao, Tang, Hu, & Liu, 2015), online bookstores like Amazon, and Apple Books (Chien, Chen, Ko, Ku, & Chan, 2015), music providers like iTunes, Spotify, and Pandora (Bu et al., 2010), and online news providers (Kompan & Bielikova, 2010).

In an example featuring the video streaming service Netflix, movies are largely categorized with metadata that uses word-based descriptions of film genre (e.g., mystery, comedy, action), intended age range (e.g., children, adults), topic (e.g., movies about friendship, super heroes, travel, family life), and other features (e.g., special effects, foreign films) (Titcomb, 2018). As Netflix does a thorough job of categorizing each film in its database, viewers can find movies in the Netflix database through text searches alone. However, the true value in Big Data approaches for content recommendation lies in the ability to capture and analyze a user’s viewing patterns, as well as any additional information about their interests and other digital activities. Matching search queries with the metadata tags that are attached to movies is often not enough for an algorithm to infer users’ individual interests, and thus will likely provide results that may not be relevant to individual viewers within film genres. As a result of accounting for a viewers history and preferences, the Netflix recommendation system continues to improve its ability to provide hyper-specific recommendations to users through specialized categories, or “microgenres.” These microgenres can sometimes include oddly specific film categories that sometimes only have one or two films in the category, with some examples including “emotional independent drama for hopeless romantics,” “Oscar-winning visually striking films,” and “goofy dance musicals” (LaPorte, 2019).

Recent advancements in the video game industry also provide an excellent illustration of how interactive environments can be automatically adapted based on data on prior interactions. For years, this industry has accurately assumed that as players gain expertise in a game, they might become easily bored and stop playing if more challenging aspects are not made readily available (Giakoumis, Tzovaras, Moustakas, & Hassapis, 2011; Juul, 2009). To one extreme, some scholars have even compared the continual need to evolve video game play to match players’ interests as a “content arms race” (Hastings, Guha, & Stanley, 2009). This is particularly true in modern internet-based and massive multiplayer online games, in which the conditions for “winning” a game are not clearly defined. Instead, winning the game takes secondary precedence to interacting with other players and completing a series of increasingly difficult or time-consuming challenges (Granic, Lobel, & Engels, 2014; Quandt & Kroger, 2013). To address this shift in play style associated with long-term, multiplayer games, game makers have integrated increasingly adaptive and dynamic elements to games to maintain alignment with players’ expertise, interests, and play motivations.

Because they are situated solely in digital spaces, the video gaming industry can also easily collect and use player data to refine gameplay, rules, and game interfaces (Eth-eredge, Lopes, & Bidarra, 2013). By using data on users’ previous play in real time, the current generation of multiplayer video games have increasingly provided dynamic ongoing support, delivered timely hints as players complete activities, and enabled the difficulty of games to fluctuate based on players’ demonstrated level of skill (Tomai, Salazar, & Salinas, 2012; Yannakakis & Togelius, 2011). To meet the equally important need of retaining player engagement, the use of Big Data has likewise afforded the video game industry with many new tools for adapting gameplay based on each player’s demonstrated capabilities, interests, and play styles (Pedersen, Togelius, & Yannakakis, 2010; Tomai, 2012). With these advances in mind, modern video games reveal key design parallels to researchers and instructional designers who are interested in implementing interventions for teaching specific skills and strategies to learners through adaptive learning environments.

Modern adaptive games create dynamic changes to interactive capabilities and play structure in multiple ways. The most prominent adaptive feature that game designers implement is adaptive difficulty. Based on the interaction patterns of players, adaptive difficulty games fine-tune challenges of in-game opponents, puzzles, and quests (Dziedzic, 2016; Hunicke, 2005). Such difficulty adjustments are made when interaction patterns reveal a player is stuck or not progressing, or alternatively, progressing too quickly without any challenge (Jennings-Teats, Smith, & Wardrip-Fruin, 2010;

Missura & Gärtner, 2009). These challenge-based adaptations have also included changing the types of options available during puzzle solving (Lavoue, Monterrat, Desmarais, & George, 2018), manipulating the types of actions or quests available to players based on their demonstrated skill with the game (Yannakakis, 2012), and limiting the access to locations within complex multiplayer worlds to make decisions easier for new players as they develop skill with the game (Yan & Natkin, 2011).

Another good example is a recent spur of game research that has demonstrated the success of delivering differentiated hints and player supports based on prior play data. For instance, Bommanapally, Subramaniam, Chundi, and Parakh (2018) showed that hints can be custom provided as players demonstrate difficulty or lack of progress with navigating multiplayer game worlds and with game interfaces alike. In another example by Chen and Lei (2006), they acknowledge that interacting within a multiplayer environment with many options can be difficult for new players. Thus, providing hints at critical moments can help focus and enable students to continue play. Similarly, Kang, Kim, and Kim (2010) argued that real-time analytics can drive adaptations to the non-player characters and “artificial intelligence” elements of a game, with nonplayer characters being able to adjust their dialogue to provide hints based on players’ demonstrated skill and current state in the game. As an excellent example of scaffolding, the adaptation of hints is closely aligned with the continual adjustment of players’ zones of proximal development and can continually support the development of student skills and strategies (Wauck 8< Fu, 2017).

This practice of promoting continual support and encouraging player motivation by games provides important insights to educators. The long-term process of learning is often emphasized over simple task completion, reaching the end of a curriculum, or taking tests. An emphasis on supporting the process of learning and continual motivation of students is especially true in open-ended learning activities, such as problem-and project-based learning where the “end” state of the learning experience is not necessarily known when students and teachers begin the process (Ertmer & Simons, 2006). Finding ways to automate data collection for dynamic adaption of learning activities, student supports and help mechanisms, and motivational encouragement presents an achievable challenge for educational technology design in the near future.

 
<<   CONTENTS   >>

Related topics