Student performance and moral decision making in predictive technologies

Pattern recognition is a well established technique utilized by neural networks in applications that attempt to make predictions on human behaviour within a limited context. ‘Many successful researchers have used Neural Network and Decision Trees in the subject of prediction and decision making’1. In the field of education, the future behaviour of a student can be said to be important to know in order to illicit an intervention, pre-emptively by the teacher or administration. ‘Predictive’ profiling is understood to be ‘identifying who an individual is, classifying what they are and evaluating what they might be’2. In [1], the proposed modelling approach includes a feed forward network (to predict future student performance results with past data) and an additional recurrent network that maps the current state of the student’s performance to a future desired state so that an intervention, or a decision can be made. In 3, a traditional feed forward NN was used to analyze learning behaviour (represented as data collected by a web-based system), to ‘classify students according to their usual performance and finally to predict their final grades’. The utilization of the NN technique improved classification which elevated the accuracy rate of the model to reach 89.96%.

One of the constraints of predictive technologies such as ANN is its dependency on reducing knowledge to a data-based representation[2]. This is highlighted by the challenges that neural networks face when used to solve complex moral problems. This requirement of ANN, to reduce a specific type of knowledge to a quantifiable abstraction, avoids important meta-ethical questions like what moral knowledge is and how it is even possible to have moral knowledge4. Without proof, any assumptions made invite doubt. The accuracy of what is being predicted, like what is a future moral behaviour based on past moral behaviours, will be affected by how well it can be represented by data.

Even if it is the case that moral knowledge can be accurately represented by data, it is still subject to bias. In order to classify moral and immoral actions which will be used as parameters for inputs to neural nets, decisions need to be made around what those are.  Creating a classifier is inherently a value judgement and in [4] a ‘catholic theory of ethics’ was used in part to define an ethical classifier.  The accuracy of what is being predicted needs to account for how much bias will influence the outcome.

Creative Commons License
Student performance and moral decision making in predictive technologies by Brad is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.


  1. bin Mat, U.; Buniyamin, N.; Arsad, P.M.; Kassim, R., “An overview of using academic analytics to predict and improve students’ achievement: A proposed proactive intelligent intervention,” in Engineering Education (ICEED), 2013 IEEE 5th Conference on , vol., no., pp.126-130, 4-5 Dec. 2013 DOI: 10.1109/ICEED.2013.6908316
  2. Neil Selwyn, “Data entry: towards the critical study of digital data and education,” in Learning, Media and Technology, 40:1, 64-82, 2015 DOI: 10.1080/17439884.2014.921628
  3. Jieqiong Zheng; Zeyu Chen; Changjun Zhou, “Applying NN-based data mining to learning performance assessment,” in Conference Anthology, IEEE , vol., no., pp.1-5, 1-8 Jan. 2013 DOI: 10.1109/ANTHOLOGY.2013.6784924
  4. Honarvar, Ali Reza; Nasser Ghasem-Aghaee,  “An artificial neural network approach for creating an ethical artificial agent,” in Computational Intelligence in Robotics and Automation (CIRA), 2009 IEEE International Symposium on, 290 – 295, DOI: 10.1109/CIRA.2009.5423190


I am a developer with an 18-year career in the Information Technology sector. Over the last half-decade, I’ve dedicated myself to advancing my expertise in the realm of intelligent information systems with a Master of Science in Information Systems (MScIS) degree. Notably, I recently completed a substantial socio-technical study, examining the feasibility of implementing responsible AI (RAI) within the public sector. Prior to my role in the public service, I undertook diverse software development roles as a contractor, a team lead and providing valuable services to post-secondary institutions. My driving passion revolves around the convergence of technology and the law, with a particular focus on the capacity of ethical AI systems to shed light on critical issues.

You may also like...

Leave a Reply

Your email address will not be published. Required fields are marked *