What Is Human Behavior Forecasting?
Human behavior forecasting is not a new concept. Facebook predicts what ads you are more likely to click on. Google suggests what you might be searching for. Netflix tells you what other shows you might like to watch. All of them are using the power of artificial intelligence to predict individual's needs and preferences.
But what if you could forecast business results based on an entire market trend or entire population's behavior? What if you could simulate how people are likely to respond to an idea? That's what we do here at Concentric. We use human behavior forecasting to help organizations make smarter business decisions by forecasting an entire market's response.
We define human behavior forecasting as the endeavor to anticipate (within some margin of error) the choices that people make. Choices are what products to buy, what candidate to vote for, what leisure activities to engage in, what mode of transportation to use, etc. Whenever people are faced with multiple alternatives differentiated through multiple attributes, human behavior forecasting applies.
The method of human behavior forecasting typically starts with a set of hypotheses from behavioral economics, complexity science, and network science governing these behaviors and interactions. It then trains the model against historical data until the outcomes match real-world circumstances. Because of its deductive foundation, human behavior forecasting offers a lot of explanatory power and is capable of answering many “what if” scenarios. Our whitepaper explains the recent breakthroughs that have made this analysis possible.
How does human behavior forecasting differ from other types of forecasting?
Other forecasting methods, like macroeconomic and regression models, typically make a few or all of the following simplifications:
- Populations can be reduced to one representative agent
- Agents are perfectly rational
- Feedback loops can be ignored or treated as exogenous factors
- Economic markets are priced at equilibrium
Human behavior forecasting breaks each of these assumptions.
For instance, an ARIMA model of the sales at Widget Co. -- unless the data is comprised of individual purchase histories -- simply ignores who is doing the buying. Thus, it reduces a population to one agent. If the method of human behavior forecasting was applied to this same problem it would attribute the sales to a population of buyers who are allowed different motives, behaviors, locations, friends, etc.
Recommendation engines have been wildly successful navigating the illogic and quirkiness of millions of users’ cultural tastes. Their personally tailored predictions are the precise opposite of assuming a population is homogeneous. In the beginning, these recommendations considered each choice and each user in isolation. Now the engines are incorporating social networks and internet “virality”. Since popularity is a feedback from other people’s actions affecting the next user’s choice, the engines are moving closer to the human behavior forecasting definition.
The dynamic stochastic general equilibrium model of macroeconomic forecasting assumes households and firms plan infinitely into the future and precisely optimize their long-term goals. By contrast, a human behavior forecast’s agents are allowed limited memory, limited forethought, and limited brainpower. And instead of assuming prices and other economic indicators play out to a purely logical conclusion, outcomes will emerge in the forecast as a result of all these interacting behaviors.
By incorporating more realistic detail into the model, human behavior forecasting reproduces a wider variety of phenomena than other methods. Forecasts will also maintain their validity further into the future, especially for topics involving intra-population interactions, rapidly changing industries, or path dependencies, because these effects build up over time.
When to apply the human behavior forecasting method
At Concentric, we are transforming the way companies forecast and make decisions. We’ve spent the last 15 years studying human decision-making and have built a model that integrates all the constant variables that influence human behavior. We’ve integrated this model into a forecasting & strategic planning software – Concentric Market - that allows you to forecast human behavior in a matter of minutes using the data you already have. We’ve tested its validity by answering more than 100,000 business questions across more than 20 industries.
These are just some of the most common questions our software helps answer: