Panel - Ethical and Policy Issues in Educational Data Mining
This important but difficult discussion brought together conference attendees who came from backgrounds in ed-tech industry, education, and higher-ed to discuss the future and goals of data collection, and privacy. This is certainly a topic to be discussed here in EdLab as we decide the types of analyses and research we want to conduct.
First and foremost when it comes to educational data there is the question of ownership. Data collected in the education sector is primarily about students and teachers but is collected by institutions and systems. Are the students, teachers able to download their data and specify an approved usage for their data? Can data collected by a school be easily ported to a marketing firm?
In states such as Massachusetts, legislation protects student data from any outside entity and the data is deleted as soon as the student leaves the educational system. Does this limit advancement or innovation in educational technology?
Secondly, when it comes to research there is the question of fairness, transparency and accuracy. As we have discussed in previous D&R seminars (D&R Presentation 2/14/17 STEM and Ehthics - Teaching Critical Thinking in Science) algorithmic bias is something which has great implications especially in the field of education. Would someone who is subject to an analysis, modeling, or classification be able to understand why they received that result? When you think about the topic of assessment automation do students have a right to understand why they received the grade they did when the assessment was done by a machine learning algorithm?
As researchers we must have an idea as to the purpose of our research. When creating a model are the results used for advancement or punishment? One of the panelists proposed having a section in research papers that would discuss the ethical implications of the research being conducted. What do you think of this?
The Institute for electronic engineers has as their ethical statement that their work must adhere to the following principles:
- Human benefit - Does the proposed technology benefit humans?
- Responsibility - Are the producers accountable for the technologies' design operation and legal implications?
- Transparency - Is it possible to discover how and why a decision was made by the technology
Is there information that helps the public understand the risks of applying the technology?
With some of the studies looking at click-streams or very detailed log files, there is the ability and risk of learning too much about a person. During the presentation the example of Edward Snowdens revelations about the government data collecting agency PRISM was presented. Edward Snowden was quoted as describing how your very typing process, how you type and delete words could be seen by the government. With this information you can tell how a person thinks and who they are. How do we set a limit to the data we are analyzing and should we?
In the following scenarios what do you think are the ethical considerations?
- Scenario: researcher has trained a neural network to detect individuals prone to rapid frustration and anger and applies it to a game, she releases her code on github.
- Scenario: You are building an open source system that collects assertions about students competencies from multiple sources and processes them into a competency profile that can be read through an API.
Collin Lynch - Assistant Professor of Computer Science, North Carolina State University
Danielle S. McNamara - Professor, Arizona State University
Ma. Mercedes T. Rodrigo - Professor, Ateneo de Manila University
Robby Robson - CEO, Eduworks Corporation
Yong Zhao - Foundations Distinguished Professor , University of Kansas
Neil Heffernan - Professor, Worcester Polytechnic Institute