Ethical considerations and responsible machine learning (PDF Copy)
Assessment
Please review the given paper and explain the issues with machine learning and data processing methods used in the presented work (you can write your report in bullet points if you prefer).
[Title and link to the paper will be given during the lecture session]
Notes:
Your report should be maximum 1 page (font size 11, margins 2cm from each side).
Please focus on different aspects of machine learning model(s) used in the presented work, discuss potential issues with the models and methods, and explore how these issues could be overcome. For example, whether other metrics or models could be applicable and why.
Marking Scheme:
Score range: 10%-40%
A response that explains a few relevant issues and concepts in the reviewed work but does not address the key issues or the report is incoherent or unfocused.
Score range: 41%-50%
A response that addresses some of the key issues and most of the arguments are reasonably logical. There may be some inconsistencies in the arguments. Some important aspects of the discussion may be incomplete or unconvincing, and the proposed improvements have weak propositions.
Score range: 51%-69%
A well-argued response that investigates potential issues with the proposed machine learning and data processing methods in the presented work provides a reasonable set of alternative evaluation metrics, approaches and possible improvements. The argument is relatively rational. There may be some weakness in arguments or the proposed alternative approaches and ideas, or some aspects of the modelling, processing and evaluation issues may have been overlooked.
Score range: 70%-100%
A very good response with few minor weaknesses. Potential issues in modelling, processing and evaluation of the proposed work are identified, and suitable alternatives are presented and justified. The responses use the concepts learned in this module and focus on the clinical applicability of the proposed arguments and approaches. The arguments are rational, expressed and arranged coherently, with a balanced consideration of the clinical relevance and modelling/processing complexity and applicability.
An answer that is judged to be irrelevant or missing will be given a score of 0.