Ethics and the Implications of Machine Learning
Last updated on 2026-04-01 | Edit this page
Estimated time: 15 minutes
Overview
Questions
- What are the ethical implications of using machine learning in research?
Objectives
- Consider the ethical implications of machine learning, in general, and in research.
Ethics and machine learning
As machine learning has risen in visibility, so to have concerns around the ethics of using the technology to make predictions and decisions that will affect people in everyday life. For example:
- The first death from a driverless car which failed to brake for a pedestrian.\[1\]
- Highly targeted advertising based around social media and internet usage. \[2\]
- The outcomes of elections and referenda being influenced by highly targeted social media posts. This is compounded by data being obtained without the user’s consent. \[3\]
- The widespread use of facial recognition technologies. \[4\]
- The potential for autonomous military robots to be deployed in combat. \[5\]
Problems with bias
Machine learning systems are often argued to be be fairer and more impartial in their decision-making than human beings, who are argued to be more emotional and biased, for example, when sentencing criminals or deciding if someone should be granted bail. But there are an increasing number of examples where machine learning systems have been exposed as biased due to the data they were trained on. This can occur due to the training data being unrepresentative or just under representing certain cases or groups. For example, if you were trying to automatically screen job candidates and your training data consisted only of people who were previously hired by the company, then any biases in employment processes would be reflected in the results of the machine learning.
Examples: - Amazon’s recruitment tool which was found to be biased against women - A facial recognition system which was found to be less accurate at identifying people with darker skin tones
Problems with explaining decisions
Many machine learning systems (e.g. neural networks) can’t really explain their decisions. Although the input and output are known, trying to explain why the training caused the network to behave in a certain way can be very difficult. When decisions are questioned by a human it’s difficult to provide any rationale for how a decision was arrived at.
Examples: - LLMs are generally unable to explain how they arrived at a particular answer. - Dutch SyRI system used to identify people who might be committing welfare fraud. Because the system could not explain its decisions, it was ruled to be illegal.
Problems with accuracy
No machine learning system is ever 100% accurate. Getting into the high 90s is usually considered good. But when we’re evaluating millions of data items this can translate into 100s of thousands of mis-identifications. This would be an unacceptable margin of error if the results were going to have major implications for people, such as criminal sentencing decisions or structuring debt repayments.
Examples: - The COMPAS system used to predict likelihood of re-offending by criminals in the US. - The Epic Sepsis Prediction Model used to predict likelihood of sepsis in hospital patients was found to have an ROC AUC of 0.63, which is only slightly better than random guessing (0.5).
Energy use
Many machine learning systems (especially deep learning) need vast amounts of computational power which in turn can consume vast amounts of energy. Depending on the source of that energy this might account for significant amounts of fossil fuels being burned. It is not uncommon for a modern GPU-accelerated computer to use several kilowatts of power. Running this system for one hour could easily use as much energy a typical home in the OECD would use in an entire day. Energy use can be particularly high when models are constantly being retrained or when “parameter sweeps” are done to find the best set of parameters to train with.
A 2025 article from the MIT Technology Review attempted to estimate the true carbon footprint of training and querying a large language model. It was estimated that training OpenAI’s GPT-4 model took over 50 Gigawatt hours of energy - enough to power a large city for three days. Individual queries are difficult to quantify, but estimates range from around 0.24 to 0.93 Watt-hours per query - something like turning on a TV or running a microwave for a few seconds. But with billions of users submitting hundreds of thousands of queries every minute, overall energy usage of these models is creating potential bottlenecks in global energy production.
Ethics of machine learning in research
Not all research using machine learning will have major ethical implications. Many research projects don’t directly affect the lives of other people, but this isn’t always the case.
Some questions you might want to ask yourself (and which an ethics committee might also ask you):
- Will the results of your machine learning influence a decision that will have a significant effect on a person’s life?
- Will the results of your machine learning influence a decision that will have a significant effect on an animal’s life?
- Will you be using any people to create your training data, and if so, will they have to look at any disturbing or traumatic material during the training process?
- Are there any inherent biases in the dataset(s) you’re using for training?
- How much energy will this computation use? Are there more efficient ways to get the same answer?
Exercise: Ethical implications of your own research
Split into pairs or groups of three. Think of a use case for machine learning in your research areas. What ethical implications (if any) might there be from using machine learning in your research? Write down your group’s answers in the etherpad.
- The results of machine learning reflect biases in the training and input data.
- Many machine learning algorithms can’t explain how they arrived at a decision.
- Machine learning can be used for unethical purposes.
- Consider the implications of false positives and false negatives.