"This is not a data problem": Algorithms and Power in Public Higher Education in Canada

Title of the Paper

"This is not a data problem": Algorithms and Power in Public Higher Education in Canada

Bibliographic Information

• Subject Area: Algorithm application in higher education and its socio-technical impacts
• Keywords: Human-centered computing, artificial intelligence, higher education, algorithmic decision-making, student data, fairness, learning analytics, data governance, educational inequality, institutional power

Research Background and Issues

• Identified Problems or Challenges:

  • Higher education institutions are increasingly adopting data-driven decision systems and algorithms, but their design and implementation lack a human-centered perspective, potentially leading to ethical and socio-technical issues.
  • The use of student data and algorithmic decision-making raises serious ethical concerns, such as heightened student surveillance, privacy issues, unfairness, and data misuse.
  • The societal effects of algorithm application remain underexplored, particularly in terms of how algorithms sustain or exacerbate power dynamics between students and institutions.

• Importance of the Issues:

  • Student surveillance and algorithmic judgments directly impact students' educational experiences, privacy, autonomy, and fairness.
  • Decision-making and governance in higher education institutions are being profoundly shaped by data and algorithms, which may challenge the role and objectives of education as a public service.

• Research Motivation and Related Work:

  • The authors point out that existing research often neglects the study of algorithms in higher education as complex socio-technical systems.
  • This study aims to empirically examine the application of algorithms in higher education institutions and explore their role in reshaping power structures.

Proposed Solutions

• Research Methods:

  • The authors conducted an ethnographic case study of a public college in Ontario, Canada (Centennial College), interviewing 33 stakeholders, including students, faculty, and administrators, observing data-driven learning sessions, and analyzing institutional data and reports.
  • The "ADMAPS Framework" (Algorithm Transparency Framework) was employed to analyze the design, use, and human impact of algorithms.

• Innovative Contributions:

  • The study introduces a new model—the ASP-HEI Cycle (Algorithms, Student Data, and Power in Higher Education Institutions)—to explain how algorithms reinforce institutional power through the monitoring and utilization of student data.
  • It integrates perspectives from data ethics, human-centered computing, and higher education governance to illustrate the process by which algorithms reshape power structures in education.

• Implementation Steps and Techniques:

  • Conducted an in-depth ethnographic case study using field notes, literature analysis, and open coding to process data.
  • Analyzed how algorithms are used to predict student enrollment and graduation rates, as well as to identify "high-risk" students through Early Alert Systems (EAS).
  • Examined issues of transparency and fairness in algorithm design and data processing, combining theoretical and practical approaches to explore data-driven educational strategies.

Research Outcomes

• Specific Findings:

  • Identified three major trends in data-driven practices and algorithmic decision-making in higher education institutions:
    1. Algorithms are used to allocate resources, predict student enrollment rates, and develop new programs in response to financial pressures and resource scarcity.
    2. Increased student surveillance and extensive data collection may lead to privacy breaches and loss of autonomy.
    3. Algorithmic decision-making automates teacher-student relationships, neglecting students' personalized needs.
  • Developed the ASP-HEI Cycle model, demonstrating how algorithms enhance institutional financial sustainability while maintaining and expanding existing power structures.

• Advantages Over Existing Solutions:

  • This study provides an in-depth analysis of how data and algorithms, while improving efficiency, pose potential risks to students and stakeholders, especially under opaque and unreviewed algorithm usage.
  • Offers theoretical recommendations for designing educational algorithms with a strengths-based and human-centered perspective.

• Experimental or Evaluation Results:

  • Data analysis reveals that algorithmic decision-making fails to effectively assess risks or address biases, potentially replicating and exacerbating existing inequalities.
  • Some cases confirm that institutional "data silos" significantly hinder collaboration and exacerbate conflicts.

• Limitations and Future Directions:

  • Limitations:
    1. The study does not directly evaluate biases in the data and models, focusing instead on application and social impact.
    2. The sample is limited to one Canadian college, which may restrict the generalizability of the findings to other higher education institutions.
  • Future Directions:
    1. Deconstruct and evaluate specific biases in the models.
    2. Investigate the impact of third-party commercial algorithm tools on the educational algorithm ecosystem.
    3. Examine whether the ASP-HEI model applies to other educational institutions or national contexts.