Is bias a problem in machine learning? 

Written on
byLexi Birch

In all decision-making processes, whether human or machine, bias can result in unfair outcomes. We discuss why this can be a problem for machine learning, and what we can do about it.  

Machine Learning: Machine learning (ML) is very good at capturing signals and correlations in data. Deep learning, coupled with substantial data, and lots of computation, allows us to build systems that exploit correlations in data. This behaviour allows us to model complex problems, but it also means that these models are vulnerable to bias. 

Bias: Is defined as observing results that are systematically prejudiced. If we look at why ML models might be biased, we see that the datasets used to train and test models contain biases. Bias is introduced when it reflects the biases that exist in humans. Datasets can be further biased by what data is available, or simply what the dataset creator’s frame of reference is. For example, it is very common that a dataset specialises in a particular demographic. Someone developing a speech recognition system might collect predominantly audio recordings of male voices. Our Job as machine learning practitioners is to separate the genuine signals in the data from the bias that is discriminative or unfair. 

Fairness: ML models will contain biases, but the real question is does this product treat people fairly? Models are trained to maximise accuracy over a training set and the models with the highest accuracy overall, might not have the best performance for particular subsets of the data. Often improving accuracy for a subpopulation, like facial recognition for black women, might mean slightly lower accuracy for the more frequent class, such as white men. Fairness means trying to gauge whether the models are achieving the desired tradeoff between overall accuracy and performance for all subpopulations. 

Discrimination: With machine learning systems becoming more ubiquitous in automated decision making, it is crucial that we make these systems sensitive to the type of bias that results in discrimination, especially discrimination on illegal grounds. Machine learning is already being used to make or assist decisions in the following domains of Recruiting (Screening job applicants), Banking (Credit ratings/Loan approvals), Judiciary (Recidivism risk assessments), Welfare (Welfare Benefit Eligibility), Journalism (News Recommender Systems) etc. Given the scale and impact of these industries, it is crucial that we take measures to prevent unfair discrimination. 



These experiments were run in 2017 by a researcher at the university of Washington, Rachael Tatman. She took a set of common words and collected recordings from people who marked their gender and their place of origin and compared the performance of Google’s speech recognition software. The results show robust differences in accuracy across both gender and dialect, with lower accuracy for 1) women and 2) speakers from Scotland. This finding shows the need for sociolinguistically-stratified validation of ML systems. Before we can fix the problem, we need to be able to quantify it.  


Automatic Speech Recognition (ASR) 


Rachael Tatman, Gender and Dialect Bias in YouTube’s Automatic Captions (2017) (Note that lower “word error rate” indicates better performance) 

Dealing with Bias in ML 

Because of fears that algorithms will further entrench and propagate human biases, there have been significant efforts by the Artificial Intelligence (AI) community to avoid and correct discriminatory bias in algorithms, while also making them more transparent.  

There has been an explosion of academic interest in methods for developing fair algorithms, however fewer methods have been implemented in production machine learning systems used by governments or private companies and there is little transparency about how ML decisions are made and how fair they really are. I describe three general approaches which have been deployed: 


Data transformation:

The fairness-enhancing approaches that have achieved the most practical success seem to be efforts to improve performance by adding training data, especially for underrepresented groups. IBM’s facial gender classification system was performing poorly for dark-skinned people—and dark-skinned women in particular—IBM responded with a system trained on more representative data which reportedly reduced the error rates on dark-skinned women Tenfold. 


Algorithm manipulation:

Other methods which are of intense interest in the academic field are methods which manipulate the algorithm and introduce a penalty while learning which encourages the model to reduce discrimination. It can be hard to implement in practice – especially with overlapping characteristics, like gender, age, race. 


Outcome manipulation:

Here you could adjust for known biases directly on the output of the models. 


Dealing with Bias in Aveni 

At Aveni we are putting significant effort into: understanding, tracking and mitigating bias. 


One factor which reduces our risk of exposure to bias is the fact that many of our models are geared towards assisting humans and augmenting their abilities to make decisions quickly and with maximum information. This Human+ approach means that decisions our models make are less likely to affect the fairness of the system outcome. Where models make predictions, we link the decision directly to the evidence we had for making it. This makes the model as transparent as possible and further helps to mitigate problems with bias.  


The other key piece to making sure that bias in our systems does not lead to unfair outcomes is rigorous testing. We use this to continually improve and review our results to verify accuracy, precision and recall on different subgroups, and we do this on a regular basis. We also produce annual bias reports, which can be used for auditing and reporting purposes.  


In circumstances where we have access to large enough datasets, we deploy fairness toolkits to investigate and mitigate bias in our datasets and in our models. We are continuously evaluating the latest tools and research, to deliver the most accurate and fair models possible.  


To learn more about how our cutting-edge techniques can deliver the outcomes that you want. Visit: work with us  


You can also find us on  LinkedIn and Twitter  

Other recent posts

Is Consumer Duty just another TCF thing?

Is Consumer Duty just another TCF thing?

As the industry and analysts continue to pick apart the FCA’s final guidance on Consumer Duty, there are some skeptics wondering whether the new rules genuinely mark a shift in the regulator’s approach to...

Image by on Freepik

5 technologies to help firms comply with Consumer Duty

The Consumer Duty final guidance comes with big requirements for evidence. Firms will need to deploy advanced data-driven technology solutions to meet the regulator’s data-focused demands, ultimately adapting their approach and processes in order...

[Cybernews Interview]: Enhancing business operations with AI solutions

[Cybernews Interview]: Enhancing business operations with AI solutions

Natural Language Processing (NLP) is a field in artificial intelligence that enables machines to understand, process and analyse natural language in a similar way that humans do. As NLP becomes one of the most...

How low-resource Natural Language Processing is making Speech Analytics accessible to industry

How low-resource Natural Language Processing is making Speech Analytics accessible to industry

Natural Language Processing (NLP) is a subset of artificial intelligence that enables machines to understand, process and analyse natural language in the way that humans will. The machine analyses data, interprets, measures sentiment and...

Aveni Detect platform to revolutionise approach to Risk Assurance as Consumer Duty rules launched

Aveni Detect platform to revolutionise approach to Risk Assurance as Consumer Duty rules launched

We’re helping financial services firms meet new Consumer Duty regulations with our game-changing Aveni Detect product. Our AI platform uses the latest advances in Natural Language Processing (NLP) to monitor every customer interaction to...

How your business can reduce costs in a high inflation environment

How your business can reduce costs in a high inflation environment

Global inflation rates have been consistently on the rise over the past year. The high inflation is driven by high energy prices, issues in supply chains and an increase in consumer demand. While focus...

A woman sitting cross-legged thinking about her credit score which an indicator shows to be poor

5 crucial areas to improve to be Consumer Duty compliant

With the cost of living crisis pushing more and more people into financial vulnerability, the FCA is putting renewed pressure on lenders to provide the best possible service and support for their customers. Their...

Why a ‘Machine Line of Defence’ is critical to meeting the FCA’s Consumer Duty

Why a 'Machine Line of Defence' is critical to meeting the FCA's Consumer Duty

The FCA is constantly reminding lenders of the standards of the consumer duty of care as the cost of living keeps rising. Over the past six months, the rising cost of living has rarely been...

How Consumer Duty has made Speech Analytics Essential

How Consumer Duty has made Speech Analytics Essential

Speech analytics has been commonly used for a number of years, evolving from a novelty add-on to a powerful solution for businesses to improve their processes. With the Financial Conduct Authority’s Consumer Duty of...