Spotting the black swan: How one-off events can be predicted

Investment is inherently based on trends in the market – what has come before often can suggest what will come next.

Prediction is an important part of the process and over time this has developed into analytics, which is now supported by technology such as Artificial Intelligence (AI) and deep learning.

The result has been increased accuracy, improved mapping of how specific stocks will fluctuate and – most importantly – better investment decisions.

It’s undeniable how far investment predictions have come, but this process is far from perfect. There are still major challenges that can be found when attempting to anticipate the impact of black swan events.

These one-off, unexpected incidents have always been challenging to respond to, and although AI can better support an investor’s reaction to these changes, it is important to understand the scope of this technology.

Fundamentally, it’s impossible to anticipate the impact of something that has never happened before. There needs to be some baseline of information, details of similar incidents, or other background to help form an accurate predication that investors can put into action.

However, while black swan events like the financial crisis of 2008, Brexit, or controversial election results can seem like unique situations, there are inherent threads that can help to form a more accurate picture of how the market will react to similar events in the future. The immense complexity of drawing these threads together is clearly a mammoth challenge, but this is something that AI can solve.

Knowing the past

Past events are the foundation from which an AI algorithm can establish accurate predictions, identify trends, and analyse current market shifts. The more data to hand, the better the results from the AI.

Naturally, the data processing element that AI brings to financial institutions has been a huge leap forward in terms of how the market was analysed historically – even when other forms of fintech have been incorporated into the business.

For example, the use of AI has allowed deeper insight on trends and market motivators. Historically, even with some form of technology supporting the process, human analysis was limited to surface level findings that may not give enough insight to anticipate a black swan event or advise on particular investment choices following a shock event in the market. With AI, even disparate and seemingly irrelevant data can be drawn together to highlight certain trends which would have otherwise gone unnoticed.

This does not negate the impact of a black swan, of course – unexpected events can still have the potential to completely change the investment landscape. However, having this increased understanding of the market can be vital in modelling the repercussions of major geopolitical, financial or environmental events. When used in this way, AI can test the impacts of both real and hypothetical situations in order to offset the initial shock of a major market incident.

Seeing the potential of AI

While one-off events can be seemingly impossible to anticipate, many of the recent situations that could be classed as black swans have their roots in well understood shifts and trends in the market. Surprise election results, for example, have occurred in the past, and while each one is different, there are still common impacts that can help form a more accurate image of the market’s reaction.

As such, by having a more detailed depiction of the investment landscape, combined with a barometer on similar past incidents, the fallout from unprecedented shifts in the world can be more accurately described.

This is all the more present in the use of new AI practices such as Generative Adversarial Networks (GANs). GANs continually stress tests the market by having two separate AIs – or ‘neural networks’ – work against one another. One AI, called the ‘generator’, creates false scenarios and shows them to the other neural network, known as the ‘discriminator’.

The discriminator’s aim is to decide which information is real and which is false. With each correct conclusion, the generator learns from the discriminator’s reasoning and develops a more challenging and realistic set of data. As a result, the discriminator needs to become even better at classifying real and synthetic information, creating a learning loop where both AIs have to improve in parallel.

In the context of financial markets, this AI technology can be used to produce realistic market scenarios, and their outcome, that may not have been initially imagined. Through the generator, these scenarios could mimic the dynamics of previous black swan events as well hypothetical events that could happen in the future.

As a result, and because the AI is constantly learning, GANs have the potential to help anticipate the impact of certain market shifts for investors.

But GANs are just one of the methods AI is looking to establish a more sophisticated alternative to the well-known technique of ‘Monte Carlo simulations’, which have traditionally been used to simulate various sources of uncertainty that could affect the value of an asset or portfolio.

While these have helped with stress testing and sensitivity analysis in the past, GANs could become far more effective at producing a more accurate image of the outcome, especially for the preparation for black swan events.

Ensuring human interaction

Even with the advantages of this technology, it is also important to recognise where human engagement with these processes should come into play. Despite many concerns about AI taking over employees’ roles, in its current state, the technology is best applied to reducing the manual processes and improve the accuracy of data analysis. Put simply, there is still a fundamental need for human oversight in the process.

While GANs and other forms of AI will be able to produce market scenarios, these will be based on pre-programmed algorithms. And as the algorithm gives no flexibility when it comes to how the AI processes the data, there is always a chance that the AI produces an outcome that is either unhelpful or illogical. It is here that a person needs to be available to review the results of the AI and ensure the findings are accurate.

Black swan hunting

With the help of this human oversight, AI will become increasingly more accurate over time. The algorithm will be honed, limiting the amount of errors and inconsistencies the technology produces, allowing the results to better prepare investors for potential market shocks.

Although these situations are traditionally associated with having a negative impact on investors, with effective analysis, businesses can be better prepared to establish the right response to a wide range of scenarios.

As the saying goes, preparation is key, and this is where AI’s use for anticipating black swans can greatly help an organisation’s investment decisions. With a wide range of scenarios to work from, investors can plan suitable responses that will both minimise the potential loss and even capitalise on the situation for a bigger return.

However, to achieve this, firms need to change how they look at their data. Historically, any previous market trends and data insights have been reserved for a chosen few to analyse and implement.

As AI needs to have access to as much historical data as possible, this practice cannot exist anymore. Instead, organisations need to begin democratising the access and use of data. People in different areas of the business should have equal access to information such as trade history, investment yields and outcomes in order to better understand the decisions of the AI.

Getting started

Democratising data is not only important in order to allow the best outcome following a black swan event, it is also an important practice to maintain the reliability of AI. With greater visibility of the data, there will be an increased understanding of how the AI is interpreting the information and providing the results, meaning that businesses will be better equipped to spot potential inconsistencies, outliers and simple errors before they result in an incorrect finding.

This is usually a major hurdle for effective AI implementation. Data sets can often be inconsistent or follow various different formats. Consolidating all this information can be a time-consuming process that requires the involvement of senior individuals to establish clear guidelines on data processing, as well as more junior members to make sure that all the information follows that template.

Despite the investment needed to make company data consistent, the payoff will be a process that can be far more reliable than any human analysis or existing product being used in the business.

With the changing global landscape, it is inevitable that the financial market will experience black swan events that can completely change the value and interest in assets, industries and trends.

AI has huge potential to manage this impact, through its ability to access, digest and respond to vast amounts of data that no single human would be capable of processing. Plus, its ability to see the ‘deeper level’ trends that may not be obvious only adds to its advantages in the market.

However, to truly make black swan events more manageable in the industry, there needs to be an increased emphasis on human involvement – to sense check the findings, ensure data is consistent and allow the information to be access by a broader range of staff to ensure consistency and reliability across the board. This is where the true power of AI lies.

Comments are closed.