Director of Innovation @ Georgia-Pacific LLC (Retired)
Founder: Advanced Innovation Management |Strategic Planning Expert | Keynote Speaker | Business & Technical Solutions Advisor | Board Advisor
March 4. 2024
In the next few articles, we’ll discuss Causal AI, how it differs from other forms of AI, and its potential impact on our businesses. In this first article, we delve into the basics of Causal AI, its foundational principles, and how it differs from other AI technologies. We’ll explore how Causal AI leverages both knowledge and data to make informed decisions and learn about its potential to revolutionize decision-making processes across industries.
First, some background: Our team has accumulated more than four years of experience working with Causal AI, and the insights we've gained are invaluable for businesses seeking to stay ahead in today's competitive landscape. This journey has underscored a crucial truth: to thrive in the digital age, organizations must do more than merely keep up with technology; they must harness the full potential of Artificial Intelligence. According to Dr. Arthur Kordon, author of Applying Data Science: How to Create Value from Artificial Intelligence, true applied Artificial Intelligence should possess three essential attributes:
The ability to learn, possessing the capacity to be taught by both others and itself, just as humans do.
The ability to predict potential outcomes and discover counterfactuals based on its combined acquired knowledge.
The ability to reason and make decisions based on that reasoning for a specific purpose.
Imagine capturing the expertise of your best Subject Matter Experts (SMEs) and making that knowledge, along with relevant data, readily available for both new and existing employees. It’s not Generative AI, although we use Gen AI for communications, it’s a completely different paradigm. In this first article, we discuss the basics of Causal AI and how it employs cause and effect to make informed autonomous decisions or recommendations. In subsequent articles, we will explore how this capability could positively impact our entire workforce.
An Overview of The Basics of Causal AI:
Picture having a magic crystal ball that predicts the future based on your actions. Causal AI operates similarly to this crystal ball, but instead of magic, it relies on sophisticated algorithms and computational models to unravel cause-and-effect relationships.
Causal AI operates like a wise advisor or an agent, helping us to understand why things happen and how to achieve our desired outcomes. There are various approaches to developing Causal AI, but the method we employ involves integrating different components that work in tandem: Knowledge AI, which encompasses the rules, theories and processes of our business, and Data AI, which analyzes datasets to identify patterns and correlations. These components are built from causal models, which constitute the foundation of Causal AI. This technology assists us in making informed decisions based on both our existing knowledge and insights gleaned from data analysis.
Causal AI owes much of its development to the pioneering work of Judea Pearl, whose contributions over several decades relating to Causal Reasoning, Bayesian Networks, and Graphical Models have significantly shaped the evolution of Causal AI.
Having established the foundational principles and components of Causal AI, the question arises: how does this innovative technology translate into tangible benefits for our business?
Leveraging The Power of Data:
Currently, Data Analytics or Data AI is widely used by many of us and by companies employing teams of analysts. These analysts use data to guide decision-making, relying on correlations and probabilities. Data AI operates like a detective, scrutinizing data sets to uncover patterns and connections, offering insights into customer behavior, market trends, and other business-influencing factors. Without getting too techy technical, in Causal AI, this data forms the backbone of what is called a Structural Causal Model (SCM), it shows us what is actually happening through data.
Consider a manufacturing company aiming to optimize its production processes. The Structural Causal Model would primarily focus on analyzing the data related to production processes, such as machine utilization rates, raw material costs, and workforce productivity. It would identify correlations and patterns within this data to uncover potential factors influencing production output and quality, such as machine downtime due to maintenance issues.
However, obtaining a full understanding of the underlying reasons behind the identified factors and their causal relationships requires the integration of knowledge with the data.
Enhancing Data with Knowledge:
Next, let's look at the basics of Knowledge AI. This component consists of several elements, most notably the Principle Causal Model (PCM) and the Rational Causal Model (RCM). The PCM comprehends the rules, laws, and principles of our business, akin to how a college graduate understands the theoretical foundations of how a business operates. However, the graduate often lacks insight into how we specifically perform the processes and functions of our business. The graduate can be great with theory, but not so great on execution.
The Rational Causal Model addresses this gap. It encapsulates the knowledge of how our subject matter experts (SMEs) execute the processes and functions within our company, representing the intellectual property of how we do business and distinguishing us from competitors. Our SMEs play a crucial role in teaching the solution how to perform these processes, akin to how we train a new employee. Additionally, the platform continues learning with each action and interaction with users. It becomes an agent to help new and existing employees understand why certain actions may be better than others because it explains itself through its recommendations.
Having both the Principle Causal Model and The Rational Causal Model working in tandem is like having industry experts along with experienced professionals guiding us through the complexities of our business, facilitating informed decisions and adaptability. This form of Knowledge AI shows us what is ideal through theory and execution. It enables us to make decisions based on their knowledge and expertise, supported by data. This deep understanding of business operations and industry dynamics can enhance decision-making capabilities, provide a competitive edge and enable us to adapt quickly to changes and seize opportunities as they arise.
Making Decisions Based on Knowledge, Data, and Causality
By combining Knowledge AI, Data AI and the understanding of cause and effect through Causal AI, we create Automated Reasoning (Fig. 1.0), the first step in Special Purpose Intelligence (SPI). As Shown in (Fig 1.0), SPI functions as an intelligent agent that becomes a digital companion for our workforce. It can act autonomously or serve as a strategic advisor, providing recommendations based on an understanding of why things happen and how to achieve desired outcomes. The best part of Special Purpose Intelligence is that it is a modular capability and can be applied to different areas of the business. In the next article, we’ll dig deeper into how this agent can help in almost any domain in the enterprise.
Fig 1.0
The Pathway from Automated Reasoning to Special Purpose Intelligence
Hypotheses, Alternative Solutions, and Validations:
Causal AI not only aids decision-making but also possesses the unique ability to formulate hypotheses regarding outcomes, test those hypotheses and then validate or refute them. For instance, if we observe a decline in machine runability and raw material usage in a manufacturing plant, Causal AI may hypothesize that changes in machine health or raw materials are contributing factors.
Upon proposing a hypothesis, Causal AI doesn't halt its processes. Instead, it employs the scientific method to rigorously test and validate or disprove the hypotheses. This is also called the process of using counterfactuals (or alternative choices) and involves collecting additional problem-specific data, conducting experiments, and analyzing results in near real-time to ensure the accuracy and reliability of the proposed solution with explanations.
Systematically testing hypotheses enables Causal AI to provide deeper insights into the underlying causal relationships within our business operations. It helps us distinguish between correlation and causation, ensuring decisions are evidence-based rather than coincidental or correlated.
Through this iterative process of hypothesis generation and validation, Causal AI empowers us to make more informed decisions and optimize business processes to drive sustainable growth.
By combining Knowledge AI with Data AI, decisions become more informed. Knowledge AI looks at how things should be, while Data AI focuses on how things really are. This system uses both perspectives to find problems or gaps and fix them using cause-and-effect principles.
How Does Generative AI Fit In?
Causal AI differs from Generative AI, which encompasses language models capable of creative tasks like generating text or images, gathering and compiling data and facilitating communication. While Generative AI is essential for interacting with the Causal AI solution, the two serve distinct purposes. Causal AI primarily focuses on understanding cause-and-effect relationships through knowledge and or data to support informed and reliable decision-making. In contrast, Generative AI, such as language models, specializes in tasks like generating synthetic data and aiding communication between the Causal AI system and users.
Bringing it Together:
When we look back at the core tenets proposed by Dr. Arthur Kordon for true Artificial Intelligence, it becomes evident that Causal AI embodies all three criteria with remarkable precision. Firstly, Causal AI excels in the ability to learn, not merely by acquiring knowledge from external sources but also by continuously refining its understanding through internal processes, akin to human learning. Secondly, in terms of predicting potential outcomes and uncovering counterfactuals, Causal AI stands as a beacon of reliability, leveraging its combined knowledge and data to navigate complex scenarios with foresight and precision. Lastly, when it comes to reasoning and decision-making, Causal AI shines brightly, utilizing a sophisticated framework of cause-and-effect principles to inform purpose-driven actions and recommendations. In essence, Causal AI not only meets but exceeds the criteria set forth for true Artificial Intelligence, heralding a new era of intelligent decision-making and transformative business outcomes.
Up Next: Using Causal AI to Bridge the Knowledge Gap Left by Retiring Workers
©2025 Advanced Innovation Management. All Rights Reserved. Do Not Copy Without Written Permission.
Comments