Report: Decision-Support Systems and Human-Machine Interaction

On 15 February 2023, the DILEMA Project organised a panel discussion on Decision-Support Systems and Human-Machine Interaction at the REAIM Summit on Responsible AI in the Military Domain, hosted in the Hague by The Government of the Netherlands, and co-hosted by the Republic of Korea. The session centred around issues relating to decision-support systems and human-machine interaction with respect to applications of AI in the military domain. A multi-disciplinary group of experts discussed how the use of DSS can erode the exercise of human control, and reconfigures the role and place of human decision making and human agency. Furthermore, the session explored the implications of the discussion on DSS and HMI for the design and regulation of such systems.

Report

The report of the session is available here: Download the report

Citation: DILEMA Report, ‘Decision-Support Systems and Human-Machine Interaction’, Panel Discussion held at the REAIM Summit, 15 February 2023, Report prepared by Taylor Woodcock (2023), available at www.asser.nl/dilema/news-and-updates/report-dss-and-hmi 

Key Take-Aways

• It is crucial to focus on the interaction between humans and AI systems to properly consider the opportunities and challenges of decision-support systems.

• Decision support systems in the military context raise risks regarding the influence they have on human judgments, the speed at which these operate, challenges relating to explainability and transparency, as well questions around how to deal with system errors.

• The stages of design, development and use must be considered holistically in order to ensure effective human-machine interaction that respects legal and ethical requirements.

• International law provides applicable legal limits to the use of DSS by militaries. It is always humans that must engage in legal and ethical decision-making, AI systems may only ever support these decisions.

• When used for targeting, clarification is required on design issues for DSS around how targets are defined, what parameters are implemented, what constitutes a sufficient level of accuracy and how to deal with collateral damage that results from the use of these systems.

• Systems need to be designed, tested and monitored with the human user, relevant application and applicable laws and ethical norms in mind, with metrics in place to quantify concerns such as explainability and bias.

• Interdisciplinary teams reflecting a variety of relevant stakeholders across the development, deployment and use of algorithmic DSS should work together when algorithmic decision-support systems are embedded in military decision making.

Download the full report in PDF