Who is best at helping us to make better decisions, humans or machines? This is not an entirely straightforward question. Machines (i.e. AIs) are already making lots of micro-decisions for us every day. However, at the moment, humans are the ultimate decision-makers.

We delegate some decisions to machines (e.g. give/don’t give that person a credit card). Humans are still making decisions about what type of product or service to offer, how it is designed, marketed, sold and serviced. So what role can machines play in that?

What are the best current applications of AI?

Big data crunching
Decisions without emotion
Finding connections in disparate data sets

Big data – the rise of the machines

Artificial Intelligence has advanced hugely in the last 10 years. Quite a number of repetitive human tasks have been automated. What is much more exciting is the new capabilities that AI has opened up. For example, tools have emerged that allow intelligent brand listening and feedback monitoring of millions of discussions and reviews. These were not possible when they were human-powered.

These new AI powered tools rely on training with large datasets. That means that they are well suited to consumer facing tasks where there is lots of publicly available data.

To build useful AI machines you need accurate predictions. The problem is that in many cases, the human world is not predictable. There are general patterns that can be found in large volumes of data. These work well in general, but much less well in specific situations. If my business has relatively few customers, AI won’t help much because the dataset is too small to get accurate predictions.

Small data, humans and bias

Humans process huge amounts of “data” on a daily basis. Everything we see, hear, smell, touch, taste and imagine gets processed. More of it gets quickly forgotten because we have seen, heard, smelled, tasted or imagined it before.

All humans have mental models of the world, much like modern AI’s do. These models describe how we expect the people and things around us to behave. We experience discomfort when something challenges these models. People often describe this as a bad feeling in their gut.

These mental models can be powerful to make sense of the world. However, they are very specific to each person. They are also very prone to bias. When humans collaborate widely, these biases are often balanced out. This is why the scientific world relies so heavily on peer review.

Despite the best efforts of AI researchers, AI is usually also biased. This is usually because the data on which it was trained is biased. However, because most AI works in isolation until it is retrained, it doesn’t have the benefit of peer review.

Finding connections

Both Humans and AI are very good at connecting the dots between disparate things. Machines do this on a purely statistical basis. This approach allowed AI to identify a correlation between people who use capital letters properly on loan application forms and full repayment of the loan. The machine can only discover this if it is given the data about the application forms and the loan repayments in the first place. A human has to provide that.

Humans are much more intuitive, particularly where other humans are involved. We take contextual information into account that is very hard to build into a dataset. This includes the emotional state of another human. It includes buzz and chatter which influence human behaviour.

This means that humans are generally much better equipped to understand and empathise with humans in fluid contexts with limited data to go on. A sales person selling into a corporation has to navigate a wide range of stakeholders who are all human and subject to mood swings and personal ambition. This is really hard to automate without mass scale machine surveillance.

Machines are much better at predicting human behaviour in controlled environments. A person has put an item in their basket, what is the most likely other thing they will buy if I suggest it.

Turning knowledge into stories

Another area where humans have the edge on machines (at least for now) is storytelling. Throughout history, humans have told stories as a way of sharing knowledge. Humans learn much more through stories than raw data. This is because a story is knowledge, wrapped up in emotion. It is the emotion that engages us as humans and makes us remember things. Even the shortest stories can carry real emotional weight. By taking data and understanding its context and how it can then be applied to a given situation, you create knowledge. With knowledge, you can tell a story.

If you want to help someone to make a decision about something, tell them a story. Great storytellers understand their audience and machines are blind to their audience at the moment. They behave consistently regardless of who they are interacting with. So, when you speak to Alexa, Siri, Cortana or Google, don’t expect any special treatment.

All this means that influencing, managing stakeholders and driving significant decisions will remain a human domain for some time to come.

We might delegate many low level tasks to machines, but not the big decisions.

So how about combining the strengths of humans and machines?

Humans and Machines

What about this as a recipe…

Take the ability of:

machines to trawl large amounts of data and make predictions.
humans to empathise with people
machines to accurately predict behaviours in controlled environments
humans to connect the dots based on sparse information
machines to speed up repetitive tasks
humans to weave things together into a story that emotionally engages people

Then allow them to collaborate in the same environment, learn from each other and play to their strengths.

This is likely to produce a much better outcome in terms of decision making than either humans or machines.

The bright future is one of humans working alongside machines, with machines doing the heavy lifting and humans doing the empathising. This will allow us to make decisions that are not only data driven, but also human driven, where information has been understood and contextualised into knowledge.

To achieve this, we need platforms that enable humans and machines to work side by side more effectively. They need to be able to share data, information, insights and decisions with each other, so that they can learn from each other.