Unraveling Organizational Culture Through the Lens of Neural Networks
By: Erik Kellener
In the world of artificial intelligence (AI), neural networks (NN) have become a foundational and powerful method for machine learning. In the realm of business, organizational culture plays a pivotal role in defining and guiding the behavior of individuals and groups. On the surface, these two concepts may seem to belong to entirely different domains. However, a closer review reveals striking parallels between the two, providing critical insights for executives aiming to shape and optimize their company culture.
A Neural Network Primer
Given the vast majority of us are neither Neuroscience or Machine Learning experts, a NN primer can be useful to bridge the concepts. While there are many formal definitions of a NN, the one that seems most fitting for this primer is :
A good example to better understand a NN is to consider what a network could look like for driving a car.
[Disclaimer: This is a way oversimplified example of a NN, and may cause experts to throw up in their mouths]
At the most fundamental level, there are inputs and outputs. In this example, input nodes are the visual or sensory information (the view of the road ahead, the view from the rear, the view and side mirrors, and the speed of the car). The output nodes are how we control the car (e.g. turning the steering wheel, pressing the accelerator, or applying the brakes). Each row of nodes in the illustration is considered a “layer”.
Between the input and output, there are one or more “ hidden layers” of interconnected processing nodes. You can think of each hidden layer as a level of decision-making, similar to the thought process you go through when deciding how to react to what you see on the road. Actual NNs like the ones used for ChatGPT or Midjourney are far more complex — and can contain billions of interconnected nodes and several hundred hidden layers.
3 nodes in the middle control Parameters
Within each of the middle hidden layers is a series of parameter nodes (controls for weighting and bias) to determine how much each input (e.g driving information) contributes to the next layer (the next step in the decision-making process). For example, the weight could determine how much the view of the road ahead influences the decision to brake or accelerate, and the bias might account for personal driving style, like preferring to maintain a longer distance from the car in front.
NN can have many layers & parameters (abstracted here)
Training a NN is like the experience of learning to drive. At first, you might make mistakes, like braking too hard or not turning the steering wheel enough. But as you practice more and more (or as the NN processes more data), you get better. You learn to adjust your actions based on the inputs you receive — for instance, if you see a car ahead slowing down, and you also look in your rear view mirror for the car behind you, you know to apply the brakes to slow down while being mindful of the car behind you.
Thicker outlines indicate a trained pathway of the NN
Similarly, the NN adjusts its parameters (weights and biases) based on the difference between the actual output and the expected output. This process continues iteratively (e.g. Training) until the network’s output closely matches the expected output — in other words, until it learns to drive the car smoothly and safely.
The Parallels
Abstract and Complex
The essence of both a NN and organizational culture lies in their inherent complexity and abstractness. A NN mimics the human brain’s functionality by creating layers of interconnected neurons. Each node processes information, passing it along to others, generating an output based on collective inputs. While we can quantify aspects of neural networks, such as the weightings of different connections, the overall operation is a result of a multitude of interactions that create an intricate and complex interconnected system.
Similarly, organizational culture represents a complex and abstract system of shared assumptions, values, and beliefs, which have a strong influence on the people and the overall performance of the organization. Much like a NN, the specifics of organizational culture can be difficult to define, but its presence is clearly felt and experienced by those within it. Furthermore, we can observe and describe behaviors, norms, and attitudes, but the culture itself is an intangible entity that permeates every aspect of an organization.
Stimulus & Response
“Between stimulus and response there is a space. In that space is our power to choose our response. In our response lies our growth and our freedom.”
— Stephen Covey, Victor Frankel, or Unknown
At the core of both a NN and organizational culture lies a fundamental mechanism: stimulus and response. This dynamic process forms the foundation of how these systems operate and adapt, serving as a profound parallel between these two seemingly distinct entities.
In a NN, the stimulus corresponds to input data that is fed into the system. Using our driving example, the various driver views (front, rear, left and right) serve as the stimulus . The response is the output (steering, braking, accelerating) produced after the input data has been processed through the layers of the network. Iterating by adjusting the network parameters based on the stimulus (input) and desired response (output) allows the network to learn and improve over time.
In the context of culture, the stimulus often manifests as events, decisions, or actions within the organization. This could be a change in leadership, the introduction a new return to work policy, market shifts, or even informal cues from the behavior of colleagues and leaders. The response within a business culture is the resulting action or behavior exhibited by its members. For example, a decision by leadership to promote more open communication (stimulus) may lead to an increase in collaborative behavior (response). If the response aligns with the desired cultural values, it is often reinforced, thus encouraging similar behaviors in the future.
Optimize through Iteration
Teaching a NN involves a process of iteration and optimization. The system learns by adjusting the weightings of the network connections based on the difference between the actual and desired outputs. This is akin to how organizations shape their culture. Simon D’arcy & Todd Emaus refer to this as the Prototyping phase in their culture guide, “Scale without Losing your Soul”. This is the culture building phase where the leadership team starts living within the bounds of their defined culture code. It is an ongoing, iterative process where behaviors and attitudes are continuously influenced and aligned towards achieving a desired cultural ethos.
In both instances, there is an element of feedback that informs the system of its performance relative to the desired state. In a NN, this is accomplished by using training algorithms, while in an organization, feedback often takes the form of organizational health, performance management, surveys, and open communication channels.
Caution of Overfitting
During the training of a NN, the system learns by processing data scenarios that include both desired behaviors (e.g. braking when heading into a sharp curve in the road), and undesired behaviors (e.g. swerving to the side in the middle of a straight road). A NN that is trained too intensely on a narrow dataset becomes myopic and suffers from overfitting. The side-effect is it performs well with the exact, literal data scenarios it was trained on, but performs terrible in more generalized scenarios.
Typically organizations use cultural values as a set of street cones to help guide the business and ensure everyone is aligned in the same direction. While this can be a healthy ingredient to scaling culture, organizations sometimes can take too narrow of a view of their values, which can have a hidden darker side — Company Shadow. This is when an organization isn’t conscious or ignores the potential negative impact for adhering too literally, or narrowly to their values.
As an example, Uber’s original cultural values included “always be hustlin”, “principled confrontation” and “toe-stepping”. While on the surface, this likely helped Uber build a execution focused culture that was highly resistant to industry pressures, it had a less conscious impact that was re-enforcing an internal cut-throat culture of toxicity, and harassment as called out by the current CEO Dara Khosrowshah.
Influence
An inherent relationship lies in the influence organizational leaders and the parameters (of NN’s) have on their respective domains. In a NN, parameters carry a stronger influence on the overall behavior. These parameters are carefully adjusted during the training phase to optimize the network’s performance.
Likewise, leaders within an organization wield a significant influence on its culture. Hence the reason D’arcy advocates for a Prototype phase. Their actions, decisions, and communication set the tone for acceptable behavior and norms within the organization. By modeling desired behaviors and values, leaders can guide the evolution of the organizational culture, similar to how parameters and layers guide the learning in a NN.
Leveraging the Parallels: Insights for Executives
Understanding these parallels can equip executives with a powerful perspective on shaping their organization’s culture. Here are a few key takeaways:
Embrace Complexity: Just as a NN is complex and multifaceted, so too is organizational culture. It’s important to acknowledge this complexity and approach cultural change with a comprehensive strategy that addresses the various elements of culture.
Commit to Continuous Learning and Adaptation: Much like training a NN, shaping organizational culture is an ongoing and iterative process. Consistent feedback and adjustments are critical to aligning behavior with desired cultural outcomes.
Encourage Desired Pathways: By clearly defining, communicating and living the desired cultural code, executives can set the example and encourage behaviors that align with these values, thereby strengthening these pathways within the organization, much like the weighted pathways in a NN.
Manage the Stimulus-Response Mechanism: Executives should be aware of the stimuli they are introducing into the organization, and how these might influence employee behavior. This requires thoughtful decision-making and consistent communication to ensure the stimuli produce the desired cultural response.
Leverage Data: A NN relies on data for learning and improving. Similarly, organizations must use data from employee feedback, performance metrics, and other sources to understand their current culture and identify areas for improvement.
Promote a Growth Mindset: Just as a NN improves with training, organizations can cultivate a culture of learning and growth. This involves fostering an environment where mistakes are viewed as learning opportunities, and innovation and creativity are encouraged.
Model Desired Behaviors: Leaders play a key role in setting the cultural tone of the organization. By embodying the desired cultural code, executives can influence the organization’s cultural norms and expectations, much like how a NN’s output can be influenced by adjusting its parameters.
Intertwining NNs and organizational culture provides a unique lens for understanding and influencing the behaviors within a company. By embracing these behaviors, executives can help shape their culture. It’s a journey of continuous refinement and adaptation, but one that can lead to a resilient, cohesive, and high-performing organization.
Are you ready to unlock your organization's potential? Evolution’s coaching and consulting services are designed to help you build a thriving organizational culture and drive lasting success. Reach out to our team HERE to learn more about our bespoke offerings on Culture Development.