Cause and Effect: The Rise of Causal AI and Its World-Changing Impact
Artificial intelligence (AI) is transforming our world. You might be thinking that generative AI is the future, but the next big thing in AI is already around the corner. As a futurist, it is my job to look ahead, and I believe Causal AI is one of the most exciting areas of this rapidly-evolving field, and it will catapult AI to the next level when ready.
Causal AI refers to the use of AI to make decisions and predictions based on cause-and-effect relationships rather than just correlational relationships. This technology is gaining widespread recognition for its ability to provide more accurate insights and decision-making capabilities.
Nowadays, causal AI is being used in a variety of industries, from healthcare and finance to marketing and retail. As technology continues to advance, the development of causal AI is becoming increasingly important, because of its potential ability to learn and impact our lives. Whether you are an AI expert, a business leader, or simply someone interested in the future of technology, it is important to stay informed about the latest developments in the field of causal AI. In this article, we will explore the current state of causal AI, discussing its applications, benefits, and future potential of the next iteration of AI.
Definition of Causal AI and Its Relation to Causal Reasoning and Human Intelligence
Causal AI is a rapidly growing field that combines the power of artificial intelligence with the principles of causal reasoning. At its core, causal AI is all about understanding the underlying relationships between variables in data and using this information to make predictions and decisions.
In many ways, causal AI is similar to human intelligence in the way it thinks and processes information. Just as human beings use causal reasoning to understand the relationships between events and make decisions, causal AI uses algorithms and models to identify and analyse causal relationships in data. This allows it to make predictions and decisions based on a deep understanding of the underlying mechanisms that drive the relationships between variables.
So, what exactly is causal reasoning, and how does it relate to causal AI?
Causal reasoning is the process of understanding the relationships between causes and effects. It is the way that we, as humans, make sense of the world around us and draw conclusions based on our observations. In a similar vein, causal AI uses algorithms and models to identify and analyse causal relationships in data, allowing it to make predictions and decisions based on these relationships.
Its importance today lies in the fact that causal AI is a powerful tool that combines the power of artificial intelligence with the principles of causal reasoning to help organisations make better decisions and improve outcomes. By understanding the underlying relationships between variables in data, causal AI is able to make predictions and decisions based on a deep understanding of the underlying mechanisms that drive these relationships, much like human intelligence.
Is Causal AI Currently Adopted On a Mainstream Basis?
Causal AI is being used commercially to identify the causes of complex issues and make predictions based on past data, and relationships between variables in many companies and organisations in the healthcare, finance, and marketing industries—and then act accordingly.
However, the use of causal AI is just beginning and its use on a commercial basis is only on a small scale. Most companies are still limited in how they implement it, mostly for academic research purposes.
The companies that have used causal AI within their operations are at the service of technology giants such as Google and Microsoft, who partner with them to carry out the development of their causal AI systems and obtain insights and results on their implementation in the real world to continue improving them.
It will be within the next few years that we will see the full implementation of causal AI systems as a fundamental part of today's enterprises, just as we make use of artificial intelligence tools and platforms that have already gone mainstream and are part of our everyday lives.
Causal AI and Other Types of Artificial Intelligence
There are many different types of AI, each with its unique strengths and capabilities, and causal AI is one of the most exciting and promising of these, together with generative AI.
Unlike other types of AI, such as machine learning and deep learning, which focus on finding patterns in data, causal AI is specifically designed to identify and analyse causal relationships. For example, in healthcare, causal AI can be used to identify the underlying causes of a disease rather than just finding patterns in patient data. This can lead to more accurate diagnoses and more effective treatments.
Another key difference between causal AI and other types of AI is the way that they process information. While machine learning and deep learning rely on large amounts of data and complex algorithms to make predictions, causal AI uses a more targeted and causal approach, allowing it to make predictions and decisions based on a more nuanced understanding of the relationships between variables.
Applications of Causal AI in Business
There has been a recent explosion of interest in causality in AI. Several big tech companies have begun investing heavily in causal AI, including Microsoft, Amazon, Facebook, Google, Netflix, and Uber.
This breakthrough technology now has the potential to benefit businesses from every sector, according to industry analysts. The prestigious technology research firm Gartner described Causal AI as one of 25 emerging technologies that can transform businesses.
In order to harness causal AI's unique benefits, mature AI must be built, adapted, and adopted in the industry. The causal revolution we are witnessing right now is due to several factors.
One of the most exciting applications of causal AI in business is in marketing. By analysing customer data and identifying the underlying relationships between variables such as purchasing behaviour, demographic information, and product preferences, causal AI can help businesses to understand their customers better and target their marketing efforts more effectively. This can lead to improved customer engagement, increased sales, and higher profits.
Another important application of causal AI in business is in finance. By analysing market data and identifying the underlying relationships between variables such as stock prices, economic indicators, and market trends, causal AI can help financial institutions to make more informed investment decisions. This can lead to improved portfolio performance and higher returns for investors.
Operations and Management
In operations, causal AI can be used to optimise processes and improve efficiency. By analysing data on production processes, supply chains, and resource utilisation, causal AI can help businesses to identify bottlenecks and inefficiencies and make changes to improve performance. This can lead to lower costs, higher productivity, and increased competitiveness.
Risk Management and Fraud Detection
Causal AI is also being used in areas such as risk management and fraud detection. By
analysing large amounts of data and identifying patterns and relationships that are indicative of risk or fraud, causal AI can help businesses to mitigate these risks and protect their operations and profits.
It is important to highlight that causal AI is designed to determine the cause-and-effect relationships between variables, rather than simply identifying patterns in data. This ability allows it to make more informed predictions and inferences. Unlike traditional machine learning algorithms, which are limited to recognizing patterns, causal AI is better equipped to handle complex data and make causal connections.
The Challenges and Limitations of Causal AI
To overcome the difficulties and limitations of causal AI, it is important to remember how it operates. Causal AI relies on two main mechanisms: correlation and causation. Correlation refers to what we can observe directly, such as causal AI performing tasks. Causation refers to the underlying relationships and cause-and-effect factors that produce the data or end product, which is often a complex web of correlations.
It is still true that statistical models, even those used in advanced deep learning (DL) systems, rely on surface-level correlations to predict the future despite their notable success. Today's DL paradigm primarily emphasises maximising predictive accuracy over exploring underlying cause-and-effect relationships.
It is worthwhile to ask: What is the problem with using correlations to predict? Ultimately, we need enough predictive power in the data to make a prediction, regardless of the source of the data. The core problem lies with the brittleness of the predictions. For correlation-based predictions to remain valid, the process that generated the data needs to remain the same.
Correlation-based approaches present two fundamental challenges:
Intervening in the world is what we want to do
Most often, prediction isn't the end goal. There is often a desire to intervene in the world to achieve specific goals. Whenever we ask "What can we do to change Y?" ", we are asking a causal question about a potential intervention. We might ask: "How would increasing a loyalty incentive affect customer churn?"
Deep learning, for example, is flawed in that our actions can change the way data is generated and, therefore, the correlations we see in the data render correlation-based predictions useless in estimating the effect of interventions.
As an example, when we use a churn model (prediction) to determine whether a loyalty incentive should be offered to a customer (intervention), the incentive affects the data that was used to compute the prediction (we hope the incentive encourages the customer to stay). Causation really matters here, and we cannot simply rely on correlations to answer questions about the effects of taking a particular action (controlled experiments or causal techniques are needed).
There is a constant change in the process of generating data
Regardless of what we do, AI contexts change all the time, which makes previously useful correlations useless for prediction in the new environment. For instance, many house-price prediction AI models worked well during normal economic conditions but deteriorated as the real estate market changed.
Deep Learning models need to be more reliable and trustworthy in order to be widely implemented. In contrast to statistical correlations, causal relationships are typically more foundational and change more slowly. Using causal models, we can predict sunrise times in different locations based on variables such as the Earth's rotation and tilt with respect to the Sun (without having to hear any roosters!). It is important to note that causal models do not require this sophistication in order to be useful – humans intuitively use simple causal models on a regular basis (for instance, a ball falls when we drop it).
In machine learning, it is important to measure how well a model is working by using validation samples. But when it comes to causal models (models that look at cause-and-effect relationships), it is harder to evaluate how well they will work in the real world because it is not possible to observe what would have happened in a different scenario. This makes it challenging for businesses and organisations to use causal models and integrate them into their decision-making process.
To put it simply, it is harder to evaluate the performance of causal models and use them effectively in real-world situations.
Why is it harder to evaluate these causal models?
Because of the fundamental problem of causal inference— which refers to the difficulty in accurately determining cause-and-effect relationships. This is because, in many cases, it is not possible to observe the outcome of a scenario that did not actually occur. This is referred to as the counterfactual scenario.
For instance, let's say a causal model is used to determine the effect of a new marketing campaign on sales. To evaluate the performance of the model, it is necessary to compare the actual sales with the sales that would have occurred if the marketing campaign had not been implemented. However, since the latter scenario did not actually occur, it is impossible to observe the counterfactual outcome, making it challenging to evaluate the performance of the causal model accurately.
This raises critical research challenges in evaluating causal machine-learning models and integrating domain expertise into machine-learning pipelines. It requires a more sophisticated approach to evaluating model performance and incorporating domain knowledge, which can be a complex and time-consuming process for businesses and organisations.
Market Evolution and Future Forecast
The use of causal AI by businesses is changing as a result of several emerging trends, all of which favor the adoption of the technology. In business settings, it is increasingly recognised that AI systems must be explainable, safe, and fair. In the European Union, a proposal for new rules for AI systems has recently been inked, spurring this global trend. Under the broad-ranging incoming legislation, businesses will have to provide explainability reports, conduct stress tests, and ensure humans are kept in the loop at all times.
Why is causal AI the solution in an era where explainable, safe, and fair AI becomes more important?
One way of looking at it is that by establishing cause-and-effect relationships between variables, causal AI provides a more complete and transparent understanding of how decisions are being made. This helps to make AI more explainable, as the reasoning behind decisions is clearer.
Another approach is that by taking into account the underlying causal relationships between variables, causal AI can help to reduce the risk of unintended consequences and ensure that AI systems are safe to use, in addition to ensuring that it is unbiased and does not discriminate against any group or element of society, it also ensures that it does not discriminate against any group or element of society.
In healthcare, causal AI-enabled counterfactual analysis, or “artificial imagination”, is also becoming more prevalent. Through counterfactual analysis, causal AI achieves expert clinical accuracy in medical diagnosis. Some tasks may indeed be better suited for causal AI and result in better performance, while others may not. It is important to evaluate each specific use case and determine which approach is most suitable.
A number of high-impact use cases for social good have been examined to further investigate the efficacy of causal AI in healthcare. Among them are diagnosing childhood diseases in Pakistan and preventing women in rural India's north from avoiding hospitals. According to Harvard-affiliated researchers, we may yet find that an ounce of causal AI is worth a pound of prediction.
While AI has shown potential in certain areas, such as medical image analysis and disease diagnosis, it still faces limitations and challenges in areas such as personal communication, empathy, and ethical decision-making.
In some cases, AI can complement and support human doctors, but it cannot fully replace them. It is important to approach the integration of AI in healthcare with caution and ensure that it is used ethically and responsibly to support, rather than replace, human healthcare professionals.
Finance or Business
Many leading portfolio managers agree that causality can revolutionise current finance AI applications. Aviva Investors' Head of Investment Strategy, Michael Grady, says causal AI plays an increasingly important role in investment analysis. In addition, it allows portfolio managers and strategists to identify new causal relationships in financial, economic, and alternative data, enabling them to generate alpha—This term is used in finance to refer to the process of outperforming the market or a benchmark index, such as the S&P 500. In other words, it measures the performance of an investment relative to what would have been expected based on a benchmark.
In the context of investment analysis, causal AI is seen as important because it can help portfolio managers and strategists generate alpha by providing a more complete and accurate understanding of the relationships between variables and the factors that drive investment performance. By incorporating causal relationships into their analysis, they can make more informed decisions and potentially achieve better investment results.
In the same vein, Chris Udy, Tibra's CIO, states that causality-based techniques and automation of quantitative workflows allow us to identify more orthogonal signals faster while discarding spurious correlations.
A recent survey found that 87% of CTOs and senior data scientists agree that Causal AI can enhance their understanding of their business environment. Moreover, senior data scientists are gearing up to invest more in causality. In the future, causal inference in machine learning will be of growing importance for data-driven decision-making, and 60% plan to invest heavily in this technology.
Is there a link with Generative AI and if so, what is it?
As far as I am concerned, there is a correlation between generative AI and causal AI, as both approaches can be used for similar purposes, such as generating new data or making predictions.
Generative AI refers to a type of artificial intelligence that generates new data based on patterns learned from existing data. This can include generating images, texts, or even sounds.
Causal AI, on the other hand, focuses on understanding the relationships and cause-and-effect factors that influence the data being analysed. It can be used for prediction, causal inference, and causal impact analysis.
Both generative AI and causal AI have their own strengths and limitations and can be used in complementary ways to achieve certain goals. For example, generative AI can be used to create synthetic data for training models, while causal AI can be used to understand the causal relationships in the data and make predictions.
Causal AI can be trained on synthetic data, which is artificially created data that is meant to look like real-world data. That is to say, if the synthetic data is made in a way that accurately represents the cause-and-effect relationships between things in the real world, then the causal AI can be trained on this synthetic data and still understand the relationships.
In other words, if the synthetic data is made well or high-quality, the causal AI can understand the causal relationships in it and make predictions based on these relationships, just as it would with real-world data.
How Will Causal AI Change the Coming Years?
Looking ahead, the future of causal AI is expected to be very promising. As technology continues to advance, causal AI will likely become increasingly integrated into various industries and applications. The use of causal AI will become more widespread and sophisticated, and this will lead to a greater demand for this technology.
The market for causal AI is expected to grow rapidly in the coming years,—according to Gartner— and organisations that adopt this technology will be well-positioned to gain a competitive advantage. Since this technology is so new, there are no forecasts yet on the market size of causal AI. Whether you are in the healthcare, finance, or marketing industry, it is clear that causal AI has the potential to revolutionise the way we make decisions and predictions.
Ethical and Societal Implications
The integration of causal AI into various industries raises a number of ethical and societal implications that must be carefully considered. One such issue is bias and discrimination, as causal AI models can perpetuate existing biases present in the data and result in discriminatory outcomes. This is particularly concerning in fields like criminal justice, employment, and healthcare, where such outcomes can have serious consequences.
Imagine that a company is building a predictive model to aid in hiring decisions. The model is trained on data from the last 10 years of hiring decisions, during which the company only hired men for certain positions. As a result, the model can learn that the most important factor in hiring decisions is gender, and it can continue to make biased hiring decisions in the future, even if the company's hiring policies have changed.
In this example, the causal AI model is biased because it has been trained on biased data, and therefore may perpetuate the biases that were present in the data. This highlights the importance of ensuring that causal AI models are trained on data that accurately represent the underlying relationships and that any biases in the data are addressed before training.
Another important concern is responsibility and accountability. With the increasing complexity of AI models, it becomes more challenging to understand and explain their decisions and outcomes, leading to questions about who is responsible and accountable for these results, especially in cases where they result in harm. Privacy and security are key considerations, as causal AI often involves processing and analysing sensitive personal data.
Let’s say a healthcare organisation is using causal AI models to make diagnoses and treatment recommendations for patients; the models are trained on data from patient medical records, which contain sensitive information such as personal details, medical history, and even genetic information. If this data is not properly secured, it could be vulnerable to theft or misuse by malicious parties.
For example, the data could be stolen by hackers who then use it for identity theft or other malicious purposes. Or, the data could be sold or shared with third-party companies, who could use it for targeted marketing or other purposes without the patient's knowledge or consent. So it is important that all the data used by causal AI developers or practitioners are safe and in good hands; otherwise, the data is at risk of being misused by unauthorised individuals.
Ensuring that this data is securely and responsibly handled is crucial. Transparency and explainability are also important, as they help build trust in the technology and ensure its decisions are accountable and fair. Transparency and explainability are critical for ensuring that causal AI models are fair, trustworthy, and usable in real-world situations. They help practitioners understand how the models are making decisions and identify any potential issues or biases, and they help build confidence in the models and their results to avoid potential issues.
Finally, it is essential to ensure that AI models align with ethical values and principles in their decision-making processes and that their outcomes are just and fair.
Key Players in the Causal AI Market
From established tech giants to innovative startups, the causal AI market is home to a diverse range of companies, each with its own unique approach and vision for the future of AI. For these reasons, it is important to stay informed about the major players shaping the future of this exciting technology.
Let’s take a closer look at some of the key players in the causal AI market, exploring their backgrounds, offerings, and impact on the industry.
Google has used causal AI to improve its online advertising by better understanding the causal relationships between ad exposure and consumer behaviour. Google has also used causal AI in healthcare, where it has applied the technology to help healthcare providers make better treatment decisions for patients based on causal relationships between patient characteristics, treatments, and outcomes.
Seeking to position itself as a pioneer in artificial intelligence research, Google has been investing over the years in causal AI research, with several of its researchers contributing to the development of causal AI models and methods. Through its use of causal AI and investment in research, Google is helping to advance the field and drive innovation in the use of AI for causal inference and decision-making.
Microsoft is one of the leading pioneers in the field of causal AI and actively participates in the field to not only reap the benefits but also deliver these benefits to society by getting more precise in their research and data.
Microsoft heavily utilises causal insights to improve machine learning methods, adapts and scales causal methods to leverage large datasets, and applies all these methods to make data-driven decisions in real-world scenarios.
A great example of this is the development of DoWhy. DoWhy is an open-source Python library developed by Microsoft Research for causal inference, available on GitHub. It provides a unified framework for performing causal inference tasks, including identifying the causal relationships between variables and estimating the effects of interventions. It aims to make it easier for data scientists and machine learning practitioners to perform causal inference by providing a simple and intuitive interface for specifying causal models and performing causal analysis. It also provides a variety of tools for visualizing the results of causal analysis, including graphs, tables, and plots.
Microsoft strives to expand the use of causal methods across academia and industry through other open-source tooling and libraries, such as EconML, and Azure, and regularly presents tutorials and seminars on new methods.
The Alan Turing Institute
The Alan Turing Institute is the national institute for data science and artificial intelligence, with headquarters at the British Library in London, United Kingdom. They undertake research that tackles some of the biggest challenges in science, society, and the economy and causal AI is one of their bigger interests.
Their research is applied to real-world problems by collaborating with universities, businesses, government bodies, and not-for-profit organisations. This research benefits science, the economy, and society as a whole. They have been actively researching the subject of causal AI for the past few years, having great insights on the topic. They unveiled many realms and have extensive data on causal AI on their website, such as this publication about Counterfactual Fairness or this publication about inferential reasoning and agent-based modeling to capture relationships between individuals.
They are playing a commendable role in the ideas regarding causal AI and commercialising it.
CausaLens is a leading company in the development and implementation of causal AI. The company provides a suite of algorithms and tools for causal inference, which allows data scientists and AI practitioners to identify and analyse causal relationships in data. This enables them to build more accurate and effective AI models that can make better predictions and decisions based on these relationships.
CausaLens has a strong focus on advancing the field of causal AI and has made significant contributions to the development of new methods and algorithms for causal inference. The company has also collaborated with leading academic institutions and organisations to advance the field and apply its findings to real-world challenges.
CausaLens has proven a strong track record of delivering practical solutions for organisations that are looking to implement causal AI in their operations. The company has worked with a range of industries, including healthcare, finance, and marketing, to help organisations make better decisions and improve outcomes based on causal relationships in their data.
In terms of commercialisation, causaLens has made the most notable efforts, as in the case study with the company GameStop. This document analyses high-resolution short-interest data from 2iQ with their platform, causaLens, to gain insight into the forces driving GameStop to continue improving within the gaming industry.
With a collaboration of like-minded system designers, natural language processing experts, data scientists, and social developers, Causality Link has been able to position itself as an important player in the development of causal AI. With Causality Link's AI-powered research platform, investors and analysts gain a unique perspective on companies, industries, and macroeconomics based on the knowledge contained in millions of documents and other text-based sources.
The Causality Link platform provides clients with more telling, longer-lasting, less emotional, and more precise insights, and forecasts by aggregating explicitly stated cause-and-effect relationships between market indicators and company performance factors.
With the recent increase of artificial intelligence being applied to many different fields, the question of its real value arises. What kind of new resources will it open? How would it change everything we know and use? Wouldn't there be other ways of achieving the same results? And how safe are we to let AI do what it does best without even understanding how? These are just a few questions about artificial intelligence that remain answered for now, but perhaps, in the future, we will understand them better.
Causal AI undoubtedly presents astonishing benefits and has proven to be much more precise and sensible than conventional AI. It has opened new realms of possibilities providing a much safer gateway for present and future AI-driven tasks. To give my ending remarks, it is extremely helpful again, thanks to its cause-and-effect mechanism. It is every bit revolutionary, a more human-like AI we can (or ever have probably). The possibility of using causal artificial intelligence is likely to start new markets with an almost incalculable potential for future development.