Below is a summary of my webinar for the Learning Guild xAPI cohort 2023.
1. Understanding Data Ethics
We typically think of ethics as setting the standard of what is morally, legally, and socially acceptable and what is not. However, we can also think of ethics in a more dynamic sense. As Dr. Stephanie Moore from the University of New Mexico insightfully points out, ethics is more than a set of static principles. It’s an “ongoing reflection and conversation that articulates and deliberates, sometimes with conflicting perspectives.” In our increasingly data-driven world, we must also recognize ethic’s significance in guiding how we gather, interpret, and leverage information.
2. The Data Lifecyle & Its Ethical Implications
Every piece of data we encounter cycles through multiple stages: from its creation and curation to its processing, recording, communication, interpretation, and finally, its dissemination. This is called the data lifecycle. Ethical consideration needs to be integrated at each of the data lifecycle stage. For example, the integrity and representation of information can be profoundly influenced by the choices we make during data creation and curation. How we choose to interpret and disseminate our data findings can inadvertently place value judgments to them. As L&D professionals who interact with data regularly, we must remain vigilant and ensure that our actions align with ethical best practices throughout this lifecycle.
3. Data Ethics in L&D – Why Should You Care?
How we design, implement, and evaluate learning, there are moral implications. E-learning heavily relies on data. This data aids in tracking learner progress, personalizing content, and deriving insightful analytics. While xAPI has robust mechanism to capture, store, and retrieve intricate learning experiences, we must also address ethical considerations. It serves us help to pause, and to think about what are some of the issues we need to consider.
When we don’t take data ethics into considerations, we are likely to encounter the following:
- An incomplete picture (representation matters!)
- Unintentional consequences
- Skewed judgment and decision making ability
- The potential to cause harms
4. Why Should You Care Now?
In today’s digital age, the mere sale of location data from smartphones has blossomed into a $12 billion industry, with its legality intact in the US. This underscores the vast magnitude of the data commerce sector, hinting at its longevity and pervasive influence.
Historically, organizations have been consistent in collecting employee data, particularly for learning and performance evaluations. However, the modern twist lies in the sheer volume of data available, the diversity of learning technological platforms at our disposal, and our enhanced capability to cross-reference this data. Such triangulation offers invaluable insights, revealing patterns, informing decisions, and even predicting future trends.
As L&D involved in curating learning experiences, our responsibilities often extend to assessing and adopting new digital platforms. It’s essential to understand that these platforms come with their own set of ethical dilemmas, especially regarding data management.
A case in point is AI tools. The underlying algorithms that dictate their operations often remain shrouded in mystery. The path an algorithm takes to reach a decision isn’t always transparent, raising concerns about its accountability and explainability.
5. Challenges & Their Relevance
Privacy and Consent – xAPI can track a wide range of activities, from reading an article to attending a workshop. This granularity raises concerns about privacy, as it can potentially reveal sensitive or private information about an individual’s learning habits, preferences, strengths, and weaknesses.
Example: There is the case of the backlash on productivity scores in platforms like Microsoft 365 exemplifies the privacy concerns we face.
Ownership and Sharing – Questions about who owns the data, especially when multiple parties (e.g., content providers, learning management system vendors, and institutions) are involved, can be a challenge. If xAPI data is shared with third parties, there are potential risks of misuse. Policies need to be in place for how long xAPI data will be retained and the process by which it will be deleted.
Over-reliance on Quantitative Data – While xAPI provides a wealth of quantitative data, it’s essential to remember that not all aspects of learning can be quantified. When we base our learning interventions on quantitative data alone, we can create learning that can potentially be too prescriptive, leading to narrowed experiences.
Example: In China, there are schools that make use of a biometric headband that monitor student engagement and make sure that they are paying attention on the computer screens. The platform measures facial expressions, gestures, and voice recognition.
Bias and Discrimination – No data is free from bias. Every dataset carries the weight of its source, method of collection, and the inherent biases of those who gather it. If xAPI data is used to make decisions about learners (e.g., job promotions, academic progression, performance reviews), there’s a potential for biases in the data to lead to unfair or discriminatory decisions that favor one group or one individual over another. Such outcomes not only compromise the integrity of the decision-making process but can also have profound implications for the affected individuals.
6. Ensuring Ethical Practices
Navigating the challenges of data ethics begins with posing the right questions. Consider the following to guide your approach:
- What is the primary objective of our data collection?
- Who owns the data?
- Have we secured consents from the learner? Has this been communicated to them how we collect, store, and analyze the data?
- Are the data sources representational?
- What important metrics are missing?
- What assumptions are we making about our learners?
Once we have laid the groundwork by addressing the above questions or similar questions, our next step is to solidify our commitment to ethical standards. To achieve this, we look to three essential pillars of ethical data management:
Data Ethics Literacy: familiarize yourself with key topics around data ethics and data biases. Understand the reasons behind data collection, advocate for learner privacy, and develop a keen eye to spot and address potential biases and discrimination. It’s also beneficial to review (if there are any) your organization’s existing policies and guidelines. Check if any of your AI tools have bias-detection mechanisms, and understand how consent is sought and conveyed to learners.
Governance and Policy: Implementing a robust structure is crucial. This involves clear guidelines and accountability mechanisms. While not explicitly crafted for L&D, the 6 Guiding Principles on Data Ethics from Statistics Canada offer valuable insights that remain relevant.
Privacy by Design (PbD): Originally conceptualized for system engineering, PbD’s principles are just as pertinent to data ethics in L&D. The core idea of PbD advocates that in implementing any IT system or business practice, they need to identify and collect the minimum amount and scope of data needed (thus maximum degree of privacy) to achieve the relevant purposes.
In conclusion, as learning professionals, educators, or learners, we must take an active role in understanding and advocating for data ethics, especially in the e-learning space. I encourage each one of you to explore further, ask questions, and ensure ethical practices in your domain.