I've read a lot about the Internet of Things (IoT), but I've read very little that covers the accuracy of the endpoints. What if business decisions are made assuming their accuracy? The analytics will look good, but the raw data will be in error. I cannot confront the IoT endpoint itself, so who has the liability for errors: the endpoint manufacturer, endpoint implementer, the data analytics system, or IT staff?
This question was on my mind when I attended the Internet of Things Global Summit in Washington, D.C. earlier this month. The last session of the event was a panel addressing privacy, at which I asked the panel to address the issue of IoT measurement accuracy. Michelle De Mooy, acting director for Privacy and Data Project at the Center for Democracy and Technology responded to my question, stating that accuracy was not included in her research on IoT devices like wearables, even though she thought accuracy is extremely important.
What prompted the question was a video I saw on a segment of NBC's Rossen Reports, titled "Fitness trackers: Do they count steps, calories accurately?" In the segment, Jeff Rossen took two IoT phone applications and three wearable wrist devices, and compared them to each other and then to tests performed in the laboratory. The disparities in measuring steps and especially calories proved inaccurate. What if your doctor is using this information to base his prescribed medical treatment on, or to determine premiums for your life insurance or health insurance? The accuracy of information of the phone applications and wearables devices is questionable. According to the Rossen report:
Let's apply these measurement devices to a chair that I sit in when I'm interviewing for a job. I may not know such devices are measuring my body at the time of the interview. If I'm stressed from some other situation or expect to be stressed after the interview, my health measurements might be way off. Some might want to read the data as part of a lie detector. If the interviewer analyzes the data produced by the chair, he or she may infer certain conditions about me that are inaccurate. Some may turn me down for the job. Would this be discrimination based on data which cannot be proved to be accurate or even reproducible? Do I have any recourse in such a scenario?
What about the sensors in a vehicle measuring my driving? Do these sensors know that when the tire air pressure is increased or decreased, it does not produce the accurate speed of your vehicle? Could this data then be used to determine my fitness as a driver?
What if my stove at home is being measured from some remote location? Assume that for some reason my heating system is not working properly and I use the stove and oven as a means to heat my home. The use of the stove and oven in this way is not normal activity and sensors may not have that level of intelligence. What if I have to do this for several days in a row? Will the measuring organization be able to interpret this as abnormal behavior or will faulty conclusions be drawn?
My biggest concern about IoT is that once the data is measured, the collector, distributor, the processor, or analyzer will probably assume that it is correct. But what if it's not? How can this be remedied? It reminds me of incorrect data on a credit report. What could happen is the same as the recent situation with Wells Fargo customers who had their financial records modified and now have poor credit ratings. The customers need to spend time and money trying to correct their credit report.
I would like people to understand that you cannot assume the IoT endpoints will be right all the time. They can break. They can be modified by temperature, humidity, location, age, and other factors. I think the discussion of IoT accuracy should be raised, because we need to create policies and regulations to deal with this and provide correction mechanisms. I'm afraid that many people will be making decisions on analyzed data, but the fundamental data may inaccurate.
