This year, I've been fascinated by Herb Simon. We're on nickname terms, he and I.
There is much that Herb Simon should be known for, including:
- Founding the Business AND Computer Science programs at Carnegie Mellon... and revamping the Psychology program, too
- Elucidating the concepts of satisficing and bounded rationality (which he won a Nobel for)
- Developing "Information Processing Language," or IPL, with A Newell and JC Shaw (developing the key ideas of symbolic linked lists and the ability for a program to be written in its own language, as Lisp was ultimately based on)
- Writing, using IPL, the Logic Theorist program, the first program to solve geometric proofs and simulate human problem solving
- Participating in the 1956 Dartmouth AI conference-- although ultimately for Simon (paraphrased) "too many focus on the 'artificial' part of artificial intelligence. I'd like us to focus on the 'intelligence' part"*
- Influencing countless other thinkers, from Donella Meadows to Elinor Ostrom to Mary Catherine Bateson
"The thread of continuity," for Simon-- according to this bio from CMU-- was "his interest in human decision-making and problem-solving processes, and the implications of these processes for social institutions."
These days, many people who directly benefit from the seeds of Herb's works don't know who he is. A friend suggested I write something about *why* people today should care *more* about Herb Simon. This is not that piece. What this piece *is* though is a reflection on how my buddy Herb's thoughts and ideas have helped ME think through things over the past year.
Herb was quite prolific; you can find more of his work in the CMU archives. My favorites include: (1) By way of introduction, the interview "A Conversation with Herb Simon", (2) The Shape of Automation for Men & Management (1965), and (3) The Sciences of the Artificial (1968). An excellent biography about his approach is Herbert A Simon by Hunter Crowther-Heyck. And, while I actually haven't read it yet, his final edition of Administrative Behavior captures the organizational decision-making aspects of his work over 50 years.
What so captured my interest in Herb Simon's work? Herb found a way to purposefully combine themes of successful organizations, cross-disciplinary efforts, ways of thinking, symbolic linked lists as languages, and frameworks to address complexity-- all themes I find incredibly rich and important, but themes I have been struggling to bring together on my own. Interestingly, his initial interests-- spawned in Chicago and in Berkeley-- were in how to best organize people for better governance of cities... goals that drive me as well. For all of these reasons, Herb has become a true thought mentor of mine.
There are three Herb-influenced themes I've found particularly useful this year:
- Curated trial and error | Personal
- The theory is the program | Product / Outcomes
- Nearly decomposable systems | Organization
Each of these ideas relates to something top of mind: What is my Personal approach to the world around me, How should I construct the Product or Outcomes that I focus my efforts on, and How do I help purposefully structure the Organization that shapes those efforts? Certainly, these themes are not siloed. For example, I could say that the concept of "Nearly decomposable systems" is also a personal one (and I have, here). But for now, this categorization is useful to me as I think these ideas through.
Curious to learn more? Read on.
Curated trial and error
In his essay, "The Architecture of Complexity" (included as the last chapter in The Sciences of the Artificial), Herb writes: The process of human problem solving "ordinarily involves much trial and error"… but "the trial and error is not completely random or blind; it is in fact rather highly selective." Perhaps this is what intuition really is: selective trial and error. According to Herb, we use heuristics to suggest which paths should be tried first, which leads are most promising. We narrow the search space, guided by rules of thumb, to find a solution. Over time, as we encounter more similar patterns, we increase our speed to solve, reducing our energy to solve as well.
This concept of curated trial and error comes at a personally perplexing time for me. Do I stay a generalist, or do I specialize? When posed this way, I know the answer for myself. I'm not a fan of "specialization," a niched narrowness that doesn't extend to other applications, limiting impact in my opinion. (Of course, this could be great if your concentrated specialization has high density! But likely not the case for me.) That being said, the idea of "specialty" I'm a fan of. Specialty: a honed capability for trial and error. Specialty garnered through practice of experience. Specialty applied generally.
<Pursue curated trial and error> becomes valuable experiencing-self advice. It's a whole lot less scary than advice like <Become an [XYZ] Expert>. It also reminds me that the smartest people around me likely don't have it all figured out, either. They're continually learning and trying new things as well. I might not have a magical gift for [XYZ], but I can trust in my ability to select from the search space and quickly learn from the result. I can also trust in my judgement for which areas to garner experience in and when to read diminishing returns as indicator to try a new arena.
Trial and error is itself servomechanism theory-- error correcting theory. By using feedback loops in real time, we correct for error, the delta between our current and desired state, and we adjust our performance as necessary towards the goal. These concepts of servo theory and feedback, choice, and control-- concepts I encounter in my own work in building automation systems-- is what originally drew me to systems theory and cybernetics earlier this year, and what ultimately led me to Herb. The loop closes.
But let's not forget Herb's recommended tool for curated trial and error: heuristics.
This year, I've been searching for tools to help me better illuminate patterns. Herb Simon's own pursuit of the Logic Theorist, through heuristics demonstrated through programming languages, resonates deeply. I am not a programmer. But I do work at a software company, in the market facing side-of-house. This is a new role for me, a new discipline, and while it's in a similar industry (commercial buildings) it's completely different from my former engineering consulting days.
Seeking tips and tricks with this role switch, what I found was the bulk of business books and articles I encountered did *not* share first principles, but rather shared how a single person addressed a unique situation. These concrete examples are important, but what I needed was a grounding in foundational heuristics first. Tools to support as scaffolding while I undergo my own trial and error. As Newell, Shaw, and Simon write in their description of the Logic Theorist... "the storage of theorems has much broader transfer value than the storage of proofs."
To my delight, the heuristic of curated trial and error applies not just to my professional approach, but to my own life and dealings with all sorts of problem solving and people as well. Heuristics help us simplify in order to make sense of things. Per Simon's Nobel Prize-inducing idea-- we construct our own simplified models of real situations and then act according to *these* models as heuristic. This rationality becomes inherently bounded; I act rationally according to the simplified model I've built for myself. And that might be different from yours. (Think of the combinatorics as different rationality models meet across an org!)
The theory is the program
Hunter Crowther-Heyck writes in his biography of Simon: "Perhaps the most startling claim Simon and Newell made for their [Logic Theorist] creation was that 'the program is the theory.' They believed that in the program they had discovered the formalism appropriate to the description and explanation of the behavior of complex, adaptive systems." He goes on to state: Simon and Newell "believed that information theory, game theory, and servo theory all fell short when it came to describing human behavior. The program was the only appropriate formalism" for such a description, one that could "generate unpredictable complexity from determinant simplicity."
The theory for human problem solving, as Newell, Shaw, and Simon conclude, is the program, the "specification of what the organism will do under varying environmental circumstances" given its capabilities... not the observed behavior itself. The theory is the program.
A lesson from a phenomenal Negotiations course I took one semester echoes this point: Get people to buy in to the process. Then, even if the outcome is not what they're hoping for, they will be more likely to trust the process, and thus trust the result. Sides in a negotiation may not agree with the generated behavior, but they should agree in the program of primitive information processes to generate it.
How has this concept shaped my perspective this year? It's influenced my approach to building product. More specifically, it has taught me to not ignore the program in the quest for an outcome. For example, the ability to dogfood our own product at work has been critical to our understanding not just of the user experience of the product, but also the process by which we build and deliver it.
This is a core concept of ILP, the language Simon and Newell developed for their Logic Theorist program, and what Lisp is largely based on. Use the program to write the program. Use your product to build the product. Such an approach is similar to the way Github builds both its product and organization around repos, or how Broad Group (a Chinese company) combines its various businesses and cultural values to develop skyscraper cities. This is important in building partnerships with other companies as well: build relationship through relationship. Structure partnership agreements in ways that foster stronger partnership agreements. As I think of my future endeavors, I believe this concept will continue to deliver. Build such that your product can be used to build further product.
Nearly decomposable systems
In The Shape of Automation, Simon describes organizations as "systems of behavior designed to enable humans and their machines to accomplish goals." To Simon, the structure that organizations should take to succeed at their goals is a hierarchic one; one of stable, nested subsystems that can adapt to growing complexity.
The idea of hierarchies as an organizational framework, composed of "nearly decomposable" subsystems, has fundamentally changed the way I think about organizations over the past year. I recall a run with friend earlier this year. He was sharing his posit that teams should at best be able to operate in isolation; heightened collaboration between teams-- for productivity purposes-- is unnecessary. At the time, I totally disagreed with this outlook. I celebrated the idea of inter-team collaboration; wouldn't collaboration at every layer between teams foster productivity?
Herb's concept of nearly decomposable systems, from his essay "The Architecture of Complexity," changed my thinking. "Why?" I hear you asking. Because I was seeing my once-held view *not* working well in practice. With collaboration at every layer, my organization ended up in confusion. Communication was muddled. Ownership, berserk. In nearly decomposable systems, however, the "interactions among subsystems are weak but not negligible." Stable subsystems allow for a more adaptive whole. Then, "collecting points" of interaction (as Herb calls them in The Sciences of the Artificial) enable more decentralized decision making that still contributes to progress of the whole. The question then becomes-- what is the best level of decentralization given the organization and task environment?
As I think about how we structure the teams at my organization, and in particular, how we think of the nesting pattern and collecting points among them, I now start from the Nearly Decomposable heuristic. Are our teams stable on their own? Are dependencies collecting at the right place? This extends to the partnerships I build with other companies. Do we have the right collecting points between our organizations, suited to the task environment of our respective orgs?
Curated trial and error. The theory is the program. Nearly decomposable systems. These heuristics, care of my buddy Herb, have begun to offer me a more structured way for dealing with new situations. What's missing, I recognize, is emphasis on emotion and relationships. My sense is that this is largely missing from Herb Simon's writing and thoughts. Odd (and alarming?) for such a bulk of work focusing on human behavior. How do we understand bounded rationality within infinite games without understanding the emotion we each bring to every new situation? How do we understand our world without recognizing the sentimental nature that gives warmth to life?
Putting emphasis on a few simple tools creates compounding value. "By combinatorics on a few primitive elements, unbounded variety can be created," Herb writes in The Shape of Automation. Herb Simon has been an excellent thought mentor. Not only has he provided me with a handful of useful tools, but he also serves as a reminder that "exploring is the name of the game." In applying these heuristics across fields, from organizational design to studies of artificial intelligence, Herb demonstrated that you can have fun across multiple disciplines, in seemingly coherent fashion. And that sounds pretty great to me.
Art by Wu Guanzhong
PS. What's fun about this game, in the dance of getting to know your thought mentors, is that you ultimately are introduced to *their* thought mentors in the process. John Dewey was a thought mentor to Herb Simon, particularly his idea that learning takes place by reconciling different points of view. This is another takeaway of mine from Herb's work: as new learnings are acquired, problems are solved, and progress is advanced by reconciling the deltas.
*Funny, how ideas solidify into your memory and then you can no longer remember where you came across them. I'm still on the hunt for this citation.