Read the following vignette first to illustrate the challenge I’m making.

William and Jenny are environmentally concerned. They keep their home cooler than most people would find comfortable. When friends or family visit or stay they adjust the house temperature.

During a cold winter, an elderly relative came to stay for a week after a serious operation. They turned up the heating on the whole house. Recently their new grandchild, six months old, came to stay for the weekend with parents. Again, they warmed the whole house.

William likes cooking. On Valentine’s Day he turned up the heating in the dining room for a romantic dinner. They had local friends over to celebrate Chinese New Year. They turned up the heating in the dining room and the reception rooms.

Ok, here’s my question. Tomorrow night it will be cold (2 degrees). What temperature should the dining room be?

I think you know what questions you would ask to find out as a sentient being the optimal answer: will they have guests? Will they be entertaining?

Now let’s look at the claims over home automation.

It is argued that smart devices and AI can 'learn' and automate the decision, 'empowering' William and Jenny. This, it is argued, will facilitate the long discussed smart home of the future full of autonomous smart devices.

How exactly does a smart thermostat do that? As far as I can see from a variety of sources, it will analyse data about temperature and 'learn' William and Jenny’s preferences. Hold on a minute! The answer doesn’t lie in the temperature data, but in their calendar data!

So, if Zoe is in the diary for Friday evening, to get the right answer you’d need to know if she’s staying and who she is.
Now, for me that’s a very smart thermostat.

W Edwards Deming was one of my gurus, a father of quality systems. Google the 'read beads experiment' or the 'rules of the funnel' to see two of his examples of the misuse of data.

In statistical process control, a distinction is made between two types of variation, 'common cause' and 'special cause'. The two experiments above show the consequence on systems behaviour of not understanding the difference. The result is chaos and poor outcomes.

In the story above, the changes of temperature that William and Jenny make are 'special' in that these are changes coming from outside their routine 'system' or home.

The problem for me is that if you purely look at temperature data you will end up setting a higher average temperature than the couple would choose for themselves as all the examples above add not subtract to their preferences.

So how much value does AI add in this case? I can’t see any value at all. If you can I’d like to know.

However, if William and Jenny believe that their smart thermostat really is smart then, over time, their preferences will be determined by the thermostat. Far from empowering the smart home of the future, it will dumb down the inhabitants.

So, the AI will only work for people who have regular reliable lives to a schedule. So a software programmer working 18 hour days with no social life, working for a home automation company will see benefits.

Are you thinking what I’m thinking? Get a life...

About the author

Chris Yapp is a technology and policy futurologist. Chris has been in the IT industry since 1980. His roles have spanned Honeywell, ICL, HP, Microsoft and Capgemini. He is a Fellow of the BCS and a Fellow of the RSA.