Seesame: Ethics-based Assessment of a Sentiment Analysis Tool

Our research team Ethics and Human Values in Technology has launched a series of collaborative workshops where researchers from Seesame and KInIT had the opportunity to discuss the moral and societal consequences of sentiment analysis tools operating with real-world user data. 

Following the Assessment List for Trustworthy AI , augmented with our own approach to the assessment process, we have systematically mapped out the direct and indirect stakeholders that could be affected by the system. 

This led us to identifying various risks in domains of explainability, accuracy or human autonomy that could emerge from using sentiment analysis on social media platforms (e.g.  Facebook). During the workshops we also established the mitigation process of such situations in real life. 

Dealing with ethics in AI is however not a linear process – it calls for constant iterations, ethical considerations and consequences evaluation at every stage of the development process. 

If done correctly, ethics-based assessments can result in a set of countermeasures and actionable steps with assigned responsible roles. This way, values such as privacy and transparency are directly translated into the design of the technology and ethical risks are minimized.