The paper addresses the pervasive collection and use of personal data, termed "surveillance capitalism" by Shoshana Zuboff. It introduces situated data analysis as a method to analyze the architectures of data collection and use, emphasizing the situated nature of data construction and presentation. The author analyzes the manipulation and behavioral modification aspects of platforms dealing with personal data as environmentality, a concept developed by Jennifer Gabrys, Erich Hörl, and Mark Andrejevic, building on Foucault's concept of environmental power. Unlike disciplinary power, which relies on internalized norms, environmental power modifies the environment to promote certain behaviors and hinder others. The fitness tracking app Strava serves as a case study because users willingly provide data, which is then situated, analyzed, and displayed in various ways.

The introduction cites works by Zuboff (surveillance capitalism), Gabrys, Hörl, and Andrejevic (environmentality), Foucault (disciplinary and environmental power), and Ajana (biopolitics of quantified self). It also references existing research on self-tracking apps and their use as "technologies of the self" (Foucault), highlighting the interplay between individual self-discipline and societal pressures. The paper further draws upon work in the digital humanities that demonstrates the situated and non-objective nature of data, highlighting biases in datasets used for algorithmic prediction and machine learning.

The paper employs situated data analysis to examine four levels of situated data within Strava: (1) Personal data visualized for individual users and shared with friends and nearby users; (2) Aggregate data visualized for humans; (3) Aggregate data as an operational dataset for human users manipulating data via dashboards or (4) for machines processing data to generate new information. The analysis distinguishes between representational (data visualizations communicating to humans) and operational (data doing something through computation) uses of data. It incorporates concepts of disciplinary power (internalized norms and surveillance) and environmental power (shaping the environment to influence behavior). Analytical methods include semiotic analysis of visual interfaces, visual and rhetorical analysis of data visualizations, exploration of data provenance and omissions, ethnographic work (citing other scholars’ research), media reports, social media discussions, and technical papers describing data usage. The study also mentions potential for further methods like interviewing users or stakeholders and employing the walkthrough method and critical code studies.

Level 1 (personal data) reveals Strava as a technology of the self, where users engage in self-improvement and comparison within a local community. Visualizations foster self-discipline, aligning with societal fitness ideals. Level 2 (aggregate data visualized for humans) focuses on Strava's Global Heatmap. The analysis examines the heatmap's visual rhetoric, limitations, and biases (e.g., underrepresentation of non-Strava users, demographic imbalance). It highlights the potential for re-identification of anonymized individuals, referencing incidents where users' locations were revealed, underscoring the tension between communal data presentation and individual privacy. Level 3 (aggregate data as an operational dataset for humans) analyzes Strava Metro, where aggregated data is sold to cities. The analysis notes that despite Strava marketing it as "ground truth," the data is inherently biased due to demographic limitations. The study cites research comparing Strava data to manual cyclist counts, demonstrating the need to account for data biases. Level 4 (aggregate data for machine audiences) addresses the use of Strava data in algorithmic processes (e.g., route calculation for self-driving cars). The analysis emphasizes the operational nature of the data, its lack of direct human experience, and its role in shaping the environment (environmental power).

The findings demonstrate how the same data can be situated differently, leading to varying power dynamics. The individual level emphasizes technologies of the self and disciplinary power, while aggregate data use highlights environmental power. The analysis underscores the situated nature of knowledge, challenging the notion of objective, unbiased data. The study reveals how seemingly innocuous platforms like Strava participate in broader power structures, impacting urban planning, traffic management, and individual behavior.

The paper concludes by proposing situated data analysis as a valuable tool for understanding the complex interplay of data, power, and technology in social media platforms. It highlights the need to analyze data across different levels of usage and audience, acknowledging the inherent biases and situatedness of data. The study suggests future research exploring the application of situated data analysis to other platforms and contexts and considering the ethical implications of data usage.

The study focuses primarily on Strava, limiting its generalizability to other platforms. While the analysis incorporates multiple methods, there might be limitations in accessing certain levels of data or stakeholder perspectives. The study also acknowledges that the analysis itself is situated and potentially influenced by the researcher's perspectives.