Wiki-Conpept-Map: overview of the most important prediction methods and fields of application, analysed on the basis of English Wikipedia entries
“’Artificial intelligence’ is almost a definition of hype,” wrote Gartner, the market research and advisory company, almost a year ago, and added in February this year: “The potential is enormous, but implementation is taking longer than expected. Even with AI, big words lead to big expectations and misunderstandings – at the moment, everything that has to do with information technology seems to be AI somehow.”
It is far simpler to pose the question about the function of artificial intelligence. Put simply, AI allows better predictions to be made. The search results on Google? Predictions about my interests. Recommendations on Amazon? Likewise. Speech recognition from Siri? Ditto. So, what AI does is make predictions. This also applies to much more complex areas such as the logistics of a retailer, traffic flows or the development of a tumour.
These few examples make it clear that the possibilities of radically more accurate predictions will also radically change our lives. This raises a number of new questions: how will decision-making behaviour change in business and our private lives if we can better predict the future? What will become technically possible? And what are the most important economic and social consequences? An introduction in six parts:
Prognostics was for a long time a borderline science; predictions were often more magic than science. Now a change is taking place: more science, less fiction. That’s because, in the course of digitalisation, predictions are getting better and better. They are mainly based on the growing availability of data, the performance of computers and machine learning – a branch of artificial intelligence. Algorithms are trained to recognise empirical relationships in large amounts of data. For example, they learn to distinguish between cancerous moles and harmless freckles and achieve an accuracy that surpasses that of human experts. Such algorithms are being used increasingly often in other sectors besides medicine, such as law and the world of tax, marketing, transport and agriculture. A major difference between mathematical and magic methods is that predictions can be checked and compared in real time with real developments and competing predictions – which, in turn, improves the quality of forecasts.
Algorithms don’t confer psychic powers. They do not allow us to see into the future, but they do make facts and connections visible. Predictive analytics are task-specific prediction systems that do not aim at providing the big picture, but rather the known unknowns. The focus is on short-term or real-time forecasts, for example for weather, traffic conditions or epidemics. Known risks become more predictable, and new risk factors can be discovered earlier. With better forecasting technology, the radius in which we can safely move is widening.
First of all, the new prediction technology automates the work of experts. Prediction systems will be used for everything from financial and travel advice to nutrition and couple advice. This makes professional advice more cost-effective and available to the mass market. Smart assistants will create individual evidence-based predictions and provide recommendations for appropriate training, diet or prevention programmes. This means that even people who previously could not afford a personal consultation can consult top virtual experts at any time – in case of doubt, several at the same time. In addition, experts also receive qualified support.
Forecasting tools will also become increasingly popular in everyday life if they simplify life. Is the shopping list complete? The doctor’s appointment arranged? The laundry washed? And when do the children have to be picked up tomorrow? Such organisational questions – which clog up the brain, as it were – are called ‘mental load’. So-called ‘smart assistants’ such as Alexa or Siri could one day take over this mental load from household and family management – just as washing machines, vacuum cleaning robots and ready-made meals helped to reduce the amount of household chores.
Nobel Prize winner Daniel Kahnemann distinguishes between two types of thinking: the fast, instinctive and emotional ‘System 1’, and the slower and more rational ‘System 2’. The fast, emotional thinking of the former runs automatically. However, it tends to misjudge and make hasty decisions. The logical opinion forming process of the latter, on the other hand, must be consciously activated and is quickly utilised. Smart assistants could therefore take over its function: whenever things have to be thought through, different factors compared and alternatives examined, thinking could be delegated to the forecasting machines. Especially in environments with a lot of data (fraud detection, medical diagnoses, etc.), the machines are superior to humans. Soon, it will be inconceivable that we make important decisions without machine reinforcement.
Perhaps it will even be considered irresponsible to choose a partner, employee or politician without an algorithmic expert opinion.
Of course, there’s the danger of an increasingly infantilised society: why learn when we can have an expert mind in our pocket who knows everything better and is capable of solving all problems? On the other hand, effort, like in sports, can also be fun when it comes to thinking, and fitness for the brain could soon become as normal as it is for the body.
The spread of forecasting systems can hardly be stopped. If the competition can predict who needs new glasses and when; can forecast how many sausages a certain supermarket will sell next Saturday but one and which new employee will perform best, then you’ll go out of business if you’re only relying on gut feeling. The decisive factor for the social acceptance of forecasting technology will therefore be the issue of who exactly will become smarter with its help: all people or just a few? Forecasting machines are thinking tools that help us navigate in a hyper-complex world. This can be regarded as a good thing, in the sense that our knowledge about ourselves and the world is growing. The potential danger doesn’t lie in the increase in knowledge, but rather in the unequal distribution of knowledge. In order to limit the misuse of forecasting techniques, their use must be democratised.
To date, we tend to reflect on important decisions retrospectively: what would we change if we could start all over again? Better measuring instruments expand our imagination by visualising spaces of possibility and thus offer a laboratory in which we can simulate different development paths. So, if we had better tools, we could ask questions in advance and decide differently if we don’t like the forecasts: for another partner, another job, another apartment. Or we could avoid situations and people we’d rather never encounter. A crucial feature of a good forecasting system is that it broadens the scope and shows alternative development paths – and not that it dictates fate. Because those who know more about themselves have more opportunities to change their future. Forecasting tools help us gain new insights into our own lives based on better data, and thus better understand who we are, what motivates, stresses and makes us happy. Only those who know themselves can reduce their dependencies and expand their scope for decision-making. The inscription over the entrance to Delphi’s oracle reads: ‘Know thyself’.
Want to know more about how predictions can help your industry? Join us for the conference "The Power of Predictions – How Intelligent Machines Will Impact Decisions" on 4 June 2019. Sign up now!
How Intelligent Machines Will Impact Decisions