Trustworthiness
In this sixth blog in our Co-Creation Series, we see the relation between trustworthiness and AI in industry through our findings in the workshops on port management. The second session of the port management workshop explored the role of trustworthiness regarding AI in the port management sector. Participants ranked the most important features of trustworthiness, which were:
Accuracy (59%)
Robustness (27%)
Fairness (15%)
Accuracy
22 user requirements were found for accuracy, which were clustered into six themes:
Complex situations and responding to uncertainty
Externals factors
Automatic retraining
Enhancing accuracy and defining precision levels
Data quality
Feautures of ETA
When considering the nature of the port management sector, it is clear how accuracy is vital to the success of AI. Adhering to this requirement would make AI not just suitable for performing standard tasks, but also in long term forecasting. This can include anticipating changes and mitigating potential risks, which can be done by monitoring various global events and their potential impacts on port activities. The industry can bolster accuracy through automatic retraining and regular data update protocols; AI systems must be continuously updated and prioritise data based on its importance. The level of precision required for the completion of certain tasks may differ, and participants counselled for clarity on the preciseness that AI tools provide when using them in the port management sector.
Participants also noted the challenges posed by varying levels of technological infrastructure within respective European countries and how this may undermine the applicability of AI implementation, and so advised thorough research into exploring universally available frameworks for AI uptake. As the port management sector is highly dynamic and unpredictable, it is important that AI systems are capable of adapting to different workloads and circumstances while maintaining reliability.
Robustness
23 user requirements were found, which were clustered into six themes:
Safety measurements
Data protection
Ensuring integrity and independence
Availability
Data transparency and reliability
Data foundation and training
As discussed in last week's blog, the integrity of port systems is vital in order to prevent complete shutdown and major economic loss. With this in mind, the robustness of AI systems is critical. In order to prevent false data generation and identify potential unauthorised access, participants recommended that new AI tools should be fitted with anomaly detection systems. Furthermore, robustness can be increased by use of strong verification protocols in order to prevent sabotage or misinformation undermining the functioning of AI.
Robustness does not merely entail unauthorised influence, but also regular pressure; AI tools must be resistant to commercial influence impacting the data or algorithms if they are to be considered reliable. Moreover, participants suggested that utilising a greater range of diverse algorithms would mitigate the risk of AI tools being undermined by undue influence of individual programmers and the possibility of human error.
Fairness
A total of 26 user requirements related to fairness were generated, which were clustered into 8 groups:
Transparency and explainability
Considering port culture and commercialisation
Outside influence on the port
User responsibility and oversight
Consistent and comprehensive data
Legislation
Neutrality and equity
If AI is to strengthen the port sector, then it must be developed and implemented in such a way that all the relevant stakeholders are consulted and appreciated. Participants urged against AI monopolisation, instead recommending the use of multiple systems and ensuring that thet were mutually compatible as necessary to maintain healthy competition. Such a framework would enhance port security by spreading the risk and encourage innovation in the AI sector. Implementing AI in the port industry should be done in an evolutionary rather than a revolutionary manner, allowing time to rectify inevitable problems that may emerge in the process of such great innovation.
There was concern that financial incentives may skew the decision-making of AI tools, prioritising high-profit cargo over critical shipments like humanitarian aid; AI tools need to account for more than just economic indicators in their programming. Participants suggested that effective implementation of AI in the port management sector will require global legislation to regulate the use of such technology across borders, which will be necessary to ensure equal access to AI tools, oversee compliance and create safeguards against potential threats.
Comentarios