Maximising the efficiency of hybrid AI-decision making by ensuring transparency and trust

An AI-driven Human-centred Trustworthiness AI Optimisation System

Human genetic data can provide insights into the risks of diseases. AI has appeared to be the most effective tool for analysing complex biological data sets. However, existing AI systems lack explainability and transparency on how their results are generated which raises questions over the trustworthiness of their assessments. Professionals having trust in AI is fundamental in healthcare systems as decisions can have significant impacts.

USE CASE 1

MEDICAL UNIVERSITY OF PLOVDIV

This use-case focuses on the enhancement of datasets that are used for training of AI-based risk assessment models, with explainability and fairness criteria defined through co-creation activities in THEMIS 5.0. AI models for personalised risk assessment and prevention strategies (based in specific projections) will be optimised to make their outcome more explainable and transparent to the users. As a consequence of the AI optimisations, the healthcare professional will be empowered to make more reliable and trustworthy decisions regarding the health and cancer related risks of individuals.

Healthcare: AI Powered Personalised Risk Assessment

Ports have high import and export activity to address the demands of society and the economy. To ensure port infrastructures can support demand, timely maintenance of the docks is necessary to detect any risks and damages. If not monitored, serious risk can be posed to workers, stevedores and even visitors. AI can be used to predict disruptive events and damages that may pose a risk to those who use the port. However, many of these predictions are difficult to interpret.

USE CASE 2

PORT OF VALENCIA

The innovations developed in THEMIS 5.0 will help Port of Valencia achieve short, medium and long-term trust in the AI-driven decision support and better use the AI recommendations to address dynamic situations, resulting in potential gains in terms of time, cost, safety and user satisfaction. Through the enhancements in THEMIS 5.0 the explainable AI system will be able to provide more relevant and understandable decision support, resulting in reduced risks, costs and effort by the port authorities.

Managing Disruptions and Dynamic Situations

Social media has cemented itself as a powerful new technology that makes it easy to manipulate and fabricate information, affecting democracy and trust in society. Social networks via computational propaganda techniques, such as ‘trolling’, can amplify fake news and disinformation. Journalists risk being manipulated by actors who frequently intimidate and discredit them, and media institutions. Increasingly, AI-based tools are being used to help avoid spreading disinformation and publishing of unchecked information.

USE CASE 3

ATHENS-MACEDONIA NEWS AGENCY

In the times of ‘meta-truth’, where any decision or verdict from an AI news verification tool can be dismissed as ‘fake news’ or derided for being ‘biased’, media organisations and journalists need to be empowered with means that allows them to impose integrity standards and moral values against the socio-technical organisations needs for AI-driven news verification. This use case will allow journalists tasked with news verification to optimise the trustworthiness of the Truly Media AI services. THEMIS 5.0 will ensure that Truly Media AI services can balance the socio-technical organisational constraints of news production with the societal benefits of wider understanding thus, mitigating of the threat of disinformation.