Dealing With Biased Data
It might seem counter intuitive that although we want AI models to think like humans, we don’t want them to contain human biases. The reason being that human biases affect the models ability to be fair.
Causes of Bias
- The human factor — Machine learning’s goal is to make machines think like humans and since humans are biased the analysis of the data could lead to the machine being biased as well.
- Poor quality of training data — If the training data is biased, then the neural net will be trained with the bias.
- Model performance mismatch — Sometimes the ML training data does not match the data it will be tested on (real-world or test data) which can also lead to bias.
Types of Bias
- Algorithm bias. This is when there’s a problem within the algorithms that power the machine learning computations.
- Sample bias. This is when there’s a problem with the data used to train the machine learning model. This can be because the data used is either not large enough or representative enough to teach the system.
- Prejudice bias. This is when the data used to train the system reflects existing prejudices, stereotypes and/or faulty societal assumptions, thereby introducing those same real-world biases into the machine learning itself.
- Measurement bias. This is when the underlying problems with the accuracy of the data and how it was measured or assessed.
- Exclusion bias. This is when an important data point is left out of the data being used — something that can happen if the modellers don’t recognize the data point as consequential.
How To Minimize Bias
Choose The Correct Learning Model
Machine learning models have their own pros and cons. In a supervised model, the training data is controlled entirely by the stakeholders who prepare the dataset. While an unsupervised model depends on the neural network itself to detect bias trends. This means that the type of model must be considered when collecting the information.
Use The Right Training Dataset
Machine intelligence is only as good as its training data. The training data must be comprehensive, balanced, replicating real-world scenarios like demographic composition, and not contain humans’ biased predispositions. A good rule of thumb is to try and not reuse datasets.
Perform Data Processing Mindfully
Machine intelligence involves three types of data processing: pre-processing, in-processing, and post-processing. When you prepare datasets in pre-processing, any data that could introduce a bias should be excluded in this step. With in-processing weighting of the neural nodes must be correct to prevent a biased output. Finally, ensure there is no bias when interpreting data for human-readable consumption in the post-processing stage.
Monitor Real-world Performance Across the ML Lifecycle
No matter how carefully you choose the learning model or vet the training data, the real-world can throw up unexpected challenges. It is important to not consider any ML model as “trained” and not “finalized” and monitor them as they go.
Make Sure That There Are No Infrastructural Issues
Apart from data and the human factor, the infrastructure itself could cause bias. This is often the hardest type of bias to detect and needs careful consideration, with investment in the latest digital and technology infrastructure.
Get Eden AI’s stories in your inbox
Join Medium for free to get updates from this writer.
Are you wondering which types of biases are propagating through your AI models or how to ensure your models don’t contain these?, Send us an email and we will be happy to assist where possible. Contact us at specialists@edenai.co.za.
