When a specific technology is developed to boost efficiency, yet ultimately leads to increased energy consumption, the phenomenon is referred to as the Rebound Effect, sometimes also called the Jevons Paradox or, more prosaically, the boomerang effect. Significant efficiency improvements can produce the opposite of the desired effect if our use patterns change as a consequence. When consumers become convinced, for example, that certain products or services are particularly efficient, they might feel better about using them – which then leads to greater use. This psychological effect can be seen, for example, with more fuel-efficient automobiles, which are then driven more frequently than their inefficient predecessors. A similar phenomenon appears when the use of technological devices becomes more attractive through efficiency improvements. The processors in our laptops and smartphones have, for example, become far more efficient over the years, but partly as a result, we use more devices more frequently and for longer periods of time than ever before – in addition to loading them with a greater number of programs and apps. The result is that they (we) are consuming more total energy.

However, the Rebound Effect doesn’t just apply to end users. It can be seen in all areas where growing demand cancels out efficiency gains. Take, for example, a company that successfully integrates AI to optimize its transportation systems and sink costs. That increased efficiency may motivate the company to actually increase its overall transport volume to take advantage of the savings.

There are also situations in which putative energy savings generated through the implementation of AI systems are actually illusory – for example, when they are counteracted by higher energy costs for the server infrastructure. Similarly, longer battery life for tablets may be the result of transferring much of the energy consumption to cloud services and the data centers that power them. Energy-intensive AI models can even significantly increase the energy and resource consumption of certain apps without end users even realizing it.

The use of AI applications can also produce what’s called the “Spare Time Rebound Effect.” If, for example, a fully automated robotic vacuum mop takes over the cleaning, people have more time to do other things. But if they then fill that time with energy-intensive activities, it can lead to greater total energy consumption.

Does that mean that technological advancements in AI energy efficiency are necessarily destined to be cancelled out by the Rebound Effect? Not necessarily. Rebound Effects show that we need political and regulatory tools to eliminate them when they appear. Only then will we be able to sustainably reduce energy consumption.