Machine Learning Transparency Desperately Needed as Selection Criteria

by Nicolette Emmino

One of this year’s most popular topics among global software consultancy, ThoughtWorks, is machine learning explainability.

“Machine learning tools are used to make life-impacting decisions, however many of these models are inherently opaque. This is problematic when people need to know how a decision was made. Similarly, if the training process isn’t open, there’s a risk of introducing prejudice, sampling, algorithmic or other bias into a machine learning model,” said Dr. Rebecca Parsons, chief technology officer at ThoughtWorks.

To combat issues with machine learning and explainability, ThoughtWorkers urge business leaders and IT managers who oversee machine learning ecosystems to increase the diversity of their development teams to reduce unintentional risks in ML models, and to use tools that can reduce algorithmic bias.

This week, the company  released Volume 21 of Technology Radar, a bi-annual report based upon ThoughtWorks’ observations, conversations and frontline experiences of solving its clients’ toughest business challenges. The latest edition highlights how emerging tools such as What-If and techniques such as ethical bias testing make machine learning (ML) more intelligible, why software development should be regarded as a team sport, how to navigate the increasingly competitive cloud marketplace and the evolution towards governance as code.

Other important themes included:

  • Interpreting the Black Box of ML. Machine learning is making more decisions, but its computations are still difficult to understand. Introducing tools to increase transparency and assembling a diverse team of developers is essential to combat ML’s black box.
  • Software Development as a Team Sport. Tools and techniques that isolate members of software teams from one another hamper feedback and collaboration. Rather than focusing on individualistic memes like “10x engineers,” innovation thrives by pulling separate specializations into collaborative and cross-functional “10x teams.”
  • Cloud: Is More Less? As the major cloud providers have achieved near parity on core functionality, competition has moved to the extra services they can provide. In their haste to compete, new services are being delivered to the market with rough edges and incomplete features, so don’t assume that all services are of equal quality.
  • Protecting the Software Supply Chain. As the software development ecosystem becomes more automated, organizations should resist the ivory tower governance rules that require lengthy manual inspection and approval, and embrace automated processes.

“By discussing the blips for our Technology Radar, we identified a most precious value: knowledge (about what works well and what attempts were in vain),” says Thomas Spillecke, IT architect of cloud applications at Porsche. “The Technology Radar preserves our knowledge — but only works if we update it regularly.”

Visit to explore the interactive version or download the PDF version.

Related Articles

Leave a Comment

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More

Privacy & Cookies Policy