Hot Thought

Psychology meets philosophy: knowledge, reality, morality, meaning

Can a Thermostat Have Beliefs?

My new thermostat seems capable of important kinds of inference.

My engineer son recently bought me a Nest Learning Thermostat. The company that makes it was just purchased for more than $3 billion by Google, which is appropriately impressed by the sophistication of the functions built into the thermostat. Once the thermostat was working, I was reminded of a claim made in the 1970s by John McCarthy, one of the founders of artificial intelligence. He insisted that even a thermometer has beliefs, since it has a way of showing what the temperature is. The Nest thermostat is far more sophisticated than a thermometer, so it got me wondering whether a thermostat as complicated as this one could be capable of beliefs.

I don't think that ordinary thermometers have beliefs, because a belief operating in the human mind does a lot more than just show the current temperature. Beliefs are capable of interacting with many other beliefs through various forms of inference, including deduction, in which a conclusion follows with certainty from a set of premises, as in mathematical reasoning. People are also capable of inductive generalization, for example when they argue from observations of a new kind of food that in general that kind of food tastes good. An even more sophisticated form of reasoning is called abduction, or inference to the best explanation, in which people form and evaluate hypotheses about why things happen. Many of the most important kinds of thinking exemplify abductive reasoning: legal reasoning about whether an accused person is guilty, medical reasoning about what diseases are responsible for observed symptoms, and interpersonal reasoning about each other's mental states,

Find a Therapist

Search for a mental health professional near you.

Remarkably, the Nest thermostat seems to be capable of all of these kinds of inference. The thermostat contains the same powerful computer chip found in iPhones and iPads. Most simply, the thermostat seems capable of deductive inference when it reports the total time when the furnace was running during the day, which requires it to add up the individual times. The thermostat also has sophisticated learning mechanisms that enable it to generalize from the way that people program the temperature at different points during the day to produce generalizations of people's behaviors. The thermostat inductively altered my settings in ways that reflected my behavior. Originally, I set it for 6 AM as a starting point to come up to room temperature for the day, but Nest noticed that I was sometimes waking up earlier than six and changing the temperature. Accordingly it changed my settings to up the furnace setting at 5 AM, first on Wednesdays, then on Saturdays, and then eventually for the whole week. The thermostat recognized a change in my heating pattern before I noticed it myself.

The Next thermostat may even be capable of a kind of abductive inference, for it has motion sensors that can detect whether anyone is moving through the room in which it is placed. If it detects no movement for a while, it will infer that people are away. One way of interpreting this would be to say that the thermostat is doing an inference to the best explanation by supposing that no motion is detected because people have gone away, which it reacts to by lowering the thermostat down to a level that produces minimal heating in order to save money.

One philosophical objection to the claim the thermostat has beliefs would be that it doesn't really have representations that are about the world, what philosophers call intentionality. This objection is hard to sustain in light of the fact that the thermostat get inputs from the world in the form of its temperature and motion sensors. Moreover it produces outputs in the form of turning the heat furnace on or off and conveying information to me about its operations. Moreover I've already described how the thermostat really does seem to learn my behavior. The thermostat therefore seems to have a lot of the same functionality that justifies the attribution of beliefs to human minds and the attribution of representations to their brains.

Of course, the Nest thermostat does not have the full range of functions that human believers do, such as motor operations, consciousness, and emotions. Nevertheless, in contrast to McCarthy's thermometer, it doesn't seem crazy to me to speculate that the thermostat actually does have beliefs.

Paul Thagard is Professor of Philosophy at the University of Waterloo and author of The Brain and the Meaning of Life.


Subscribe to Hot Thought

Current Issue

Dreams of Glory

Daydreaming: How the best ideas emerge from the ether.