AI in IoT: Supporting the Ladder of Inference for better decision-making

Sibongiseni Majova, IoT Systems Developer at IoTdc.

---

Thanks to generative AI, the hype around AI in general has never been greater. What does emerging AI mean for environments that are IoT-enabled, automated and infinitely smarter than they were just 10 years ago?

We believe AI in IoT could prove to be a double-edged sword at this stage. On the one hand, it can enrich decision-making and, in time, it might be entrusted with driving strategic decisions. On the other hand, there is a risk of over-reliance on AI compounded by bias and unchallenged, bad data.

American business theorist and professor Chris Argyris described the often-invisible decision-making process as the Ladder of Inference.

The processes humans use to interpret information and make decisions as illustrated by this Ladder of Inference are strikingly similar to the way in which businesses use IoT data to strategise.

The Ladder of Inference starts with our interaction with raw data, leading us through various stages – selection, interpretation, assumption – until we finally act based on our formed beliefs. In a similar vein, IoT devices gather raw data, which companies then filter, interpret and use to make key decisions.

But in both cases, there are risks such as incomplete or incorrect data, or bias. In humans, our personal experiences and biases might influence our data selection and interpretation, possibly leading to misguided strategies. On the IoT front, companies need to ensure the collected data is accurate, secure and ethically used. As AI enters the arena, it is important to recognise that AI systems can also make biased decisions, and this too must be guarded against.

It should be recognised that no system is infallible, so human logic should be deployed to monitor and challenge data from IoT systems, particularly when there are outliers. For example, if just two of 100 sensors aren’t feeding data into a data lake, the entire system could be throwing out incorrect results. If five sensors say a fridge is at 5C, but one registers 37C, human logic dictates that this must be investigated.

Misinterpreted or mishandled data could spell disaster for a business's operations or its reputation, so it remains crucial to carefully process and interpret the data before jumping to a conclusion. And in specialised fields such as medicine, it is especially critical to have human expertise to analyse the results and make the strategic decisions.

While IoT data helps us understand customer patterns and predict trends, the Ladder of Inference encourages us to examine our cognitive biases and improve our decision-making. AI and IoT, along with human decision-making, contribute unique yet overlapping pathways to better business insights within the Ladder of Inference. Used wisely and responsibly, AI and IoT can pave the way for innovative, data-driven strategies, helping businesses thrive in our tech-driven world.

In a simple example, a garage forecourt convenience store might have a long-established system in which a technician visits the store on Fridays to check the fridges, pie ovens and food storage. If something breaks on a Monday, they won’t know about it until Friday, impacting food safety and causing perishable food losses. With the addition of a simple IoT sensor solution, the technician no longer needs to visit the store weekly – only when an appliance is faulty. This has an immediate impact on perishable losses, customer risk, salaries and costs. By adding AI and machine learning to the environment, we may start to understand failure patterns, what causes faults and how often appliances will need replacement. And human expertise enables us to ask the right questions, challenge the data and make innovative decisions to improve operations, reduce risk and boost profits.

AI has the potential to revolutionise the IoT industry, and could have unbelievable impacts in the long term. But while AI will work hand in hand with IoT in future, we cannot afford to trust it blindly at this stage. It has to be tested against human logic to ensure it is supporting unbiased, strategic and ethical decision-making.

How should we make the most of these advanced new decision-making tools while mitigating the risks?

Businesses should encourage a culture of critical thinking where assumptions can be openly challenged. And for handling IoT data, they need to adopt strict data management practices – covering everything from security to ethical use.