The foundations of the emerging data economy that are currently being laid must be aligned with environmental and climate protection goals. Yet environmental aspects have not featured substantially in proposals for the regulation of artificial intelligence (AI) to date.
This is what Peter Gailhofer and Johannes Franke show in their study on data regulation as socio-ecological agenda-setting (“Datenregulierung als sozial-ökologische Weichenstellung”) within the Ecological Research Network (Ecornet) Berlin. The two legal experts from the Oeko-Institut and the Independent Institute for Environmental Issues (UfU) have published their findings in the environmental law journal Zeitschrift für Umweltrecht (ZUR) and elsewhere, and have discussed with stakeholders from Berlin’s urban society how data on housing and building can support the city on its way to climate neutrality.
The academics have developed guiding principles that promote socio-ecological governance of behaviour-steering systems from different angles: individual and public data sovereignty, state data accountability, data transparency, data solidarity and data sufficiency.
Will digitalisation prove to be a toolbox for sustainable transformation, or an accelerant fuelling environmentally harmful patterns of growth? This will be critically influenced by the way in which we collect data and regulate its access and use in future. “As things stand, proliferation is rife”, according to the two lawyers.
Artificial intelligence makes data-based choices
If AI is fed with data it learns from that data and reinforces it. AI keeps proposing choices regardless of whether the data is the basis for good or bad behaviour patterns.
Example: Car navigation systems
In-car navigation systems are one example: they may suggest potentially slower directions to the destination to avoid a traffic jam or recommend the use of Park&Ride facilities. However, if their default setting is to propose the fastest or most convenient route, the result can be increased traffic congestion and considerably higher CO2 emissions.
Example: Wind turbine maintenance
Predictive maintenance of wind energy facilities is a second example. Status data are obtained from the systems, which then initiate upkeep work proactively and anticipatively. Depending on who supplies the data to the controlling AI, the outcomes can vary greatly: if the aim is to save resources, the service life of replaceable parts or oil changes can be optimised. If the aim is to save on maintenance costs, the ensuing environmental impacts are comparatively higher.
Example: Water consumption data
Data can generate benefits for society as a whole, but can also cause risks and harms. A utility company can use a community's water consumption data either to set profit-maximising prices or to develop strategies for reducing water consumption.