Is there any logic to outsource (our/ your) reason to AI?

Is there any logic to outsource (our/ your) reason to AI?

Logic and reason are both critical cognitive tools, but they are different and operate differently in strategic thinking and decision-making.

Logic is a systematic, structured approach to thinking that follows strict rules of inference and validity. It's about drawing conclusions based on predetermined premises using mathematical or computational principles. In a business context, logic is like a precise algorithm – it moves from A to B to C with clear, sequential steps. It's about consistency, mathematical soundness, and the ability to derive conclusions that necessarily follow from established statements. The problem is, which we will explore later, this is falling apart because logic depends on stable rules. 

Reason, by contrast, is a broader, more flexible cognitive process. It encompasses logic but embraces intuition, experience, empirical evidence, and contextual understanding. Reason allows for nuanced interpretation, considers probabilistic outcomes, and can incorporate subjective insights that pure logic will miss. Reason is the ability to make sound judgments by weighing multiple factors, understanding complex systems, and adapting to changing circumstances.

To illustrate:

  • A logical approach might say, "If all A are B, and C is an A, therefore C is B" – a perfect, mechanistic deduction because they are all A, irrespective of the label.
  • A reasonable approach would ask, "Given what we know about A, B, and C, what are the available options and course of action taking into consideration our purpose, strategic intent, potential risks, and available resources?"

In leadership, we need both. Logic provides the structural framework for sound decision-making (even in uncertain times), ensuring your strategies are internally consistent and mathematically sound (our model). Reason provides the adaptive intelligence to apply that logic creatively, ethically, and strategically in a complex, often ambiguous, volatile and uncertain business landscape.

Think of logic as the blueprint and reason as the individuals who know how to interpret and apply that blueprint to create something truly exceptional. Both logic and plans often fail. 

However, logic can fail.

Logic fails in several critical domains. These limitations reveal “logic’s” limitations as a purely algorithmic (think LLM) reasoning tool:

  1. Complex human contexts mean that logic struggles with emotional nuance, interpersonal dynamics, cultural subtleties, and situations that require incomplete or ambiguous information.

  1. In Probabilistic and Quantum Scenarios, traditional “logic” breaks down as in quantum mechanics; particles exist in multiple states simultaneously. In probabilistic systems, outcomes aren't deterministic, and scenarios have inherent uncertainty or complexity beyond binary true/false frameworks.

  1. Epistemological Limitations Logic can fail due to fundamental assumptions that might be flawed, inherent axioms that aren't universally true, Gödel's Incompleteness Theorems, which mathematically demonstrated that within any sufficiently complex logical system, there will always be true statements that cannot be proven using the system's own rules. 

  1. Real-world complexity Logic reasoning falters when confronting emergent phenomena, complex adaptive systems, situations with multiple interconnected variables and scenarios where context radically changes interpretation.

  1. In Psychology, logic cannot fully account for subjective experience, intuitive insights, creative leaps, or cognitive processes/ biases that change. 

Logic is a powerful tool, but it is fundamentally a simplified model of reasoning that breaks down when confronted with the rich, messy complexity of human experience.

Unfortunately, reason also fails.

Reason fails in equally profound and fascinating ways that complement the limitations of logic:

  1. In nearly all Cognitive Biases, reason is fundamentally compromised. Easy examples include confirmation bias (selectively accepting information that confirms pre-existing beliefs), availability heuristic (overemphasising recent or easily recalled information), Anchoring bias (becoming overly attached to initial pieces of information) and Dunning-Kruger effect (overestimating one's own competence in complex domains)
  2. In the area of emotional interference, reason can catastrophically fail when there is strong emotions cloud judgment, personal trauma or deeply held beliefs distort objective assessment, psychological defence mechanisms prevent rational analysis, or tribal or group loyalties override independent thinking.
  3. In the field of perceptual limitations, reason breaks down due to sensory limitations of human perception, neurological constraints in processing complex information, evolutionary adaptations that prioritised survival over pure rationality and limited working memory and cognitive processing capacity.
  4. In the study of Epistemological Boundaries (what is true/ truth), reason falters when confronting fundamental unknowables (the nature of consciousness, ultimate origins), paradoxes that defy straightforward rational resolution, situations with multiple equally valid but contradictory interpretations and the limits of human knowledge and understanding.
  5. Systemic and Contextual Constraints create havoc because cultural conditioning creates blind spots, institutional and social frameworks limit perspective, historical narratives constrain imaginative thinking, and technological and methodological limitations of current understanding.

A profound example is how brilliant scientists and world-class business leaders can become completely irrational when their core beliefs or professional reputations are challenged. When reason becomes a defensive mechanism rather than a tool for discovery - it fails. 

We know that logic fails due to structural limitations, and reason fails due to human psychological and perceptual complexity. Reason is a more flexible but equally fallible approach to understanding reality.

Wisdom (which is only a label for data) lies not in choosing between logic and reason but in understanding the strengths and limitations of each and cultivating the humility to recognise when our thinking and the thinking being imposed on us by those using LLMs to write might be fundamentally flawed.  Further - as a business, you need to comprehend the risk of using an LLM as its logic and reason fail. 

Outsourcing “logic” 

While economics, business models, and finance preferences logic, marketing tends to side with reason, but what is happening is that our “communication rules”, which are crafted for sales and marketing and create our income and margin from customers, are being rewritten. The rigid categories we've long used to define our world — gender, nationality, diversity are crumbling before our eyes. This isn't just an academic exercise; it's a fundamental shift that could transform how we analyse data, make decisions, understand people, and interact with emerging technologies.

For centuries, we have relied on a straightforward approach to understanding words. They mean exactly what they label (point to/ infer). "A dog is a dog," we'd say, confident in our clear-cut definitions. As the world and words have become more complex (nuanced), single words carry multiple meanings with contextual interpretations and layers.

By example. Take the word "Stud." In a traditional understanding, it means a specific category of male horse. But in reality, the meaning shifts depending on who's speaking, who's listening, and the context of the conversation - stud becomes a particular type of person. 

For the most part, this did not matter, but then entered (AI) Artificial Intelligence and (LLM) Large Language Models. The point here isn't just linguistic gymnastics; this is about how we communicate in an increasingly nuanced world - where logic starts to fail. Traditional logic treated words like fixed objects in space — immovable, unchanging. But modern communication is more like a dynamic network (reason), where meaning emerges through relationships and inference, not through rigid categorisation.

AI and LLMs aren't just fancy calculators; they're learning systems that guess the next word based on patterns and context. They "hallucinate" because they're trying to navigate a world where meaning isn't fixed but fluid and ambiguous.  The Stud is called Dave, who may be a horse or a person, the new question is does the person or horse identify as male or not? 

For business leaders, this matters profoundly. In an era of global communication, diverse teams, and rapid technological change, our old models of understanding what was said and being said are breaking down - think marketing messages/ brand identity/ brand voice but also the use of LLM deep in operations (communication). We need a more flexible approach — one that recognises that meaning comes from connection, not just definition. Think of it like a complex business ecosystem. A successful strategy isn't about rigidly defining every component but understanding how different elements interact, influence, and transform each other. The same is true for language, for technology, for human understanding.

We're moving from a world of "this is true (logic)" to a world of "this is one possibility (reason)." And in that possibility lies an incredible opportunity for some because of uncertainty - but also a new risk at a scale where your brand can be lost in a day.

Is there any danger in outsourcing logic and reason to AI?

This is a provocative and nuanced question that cuts to the heart of emerging technological and leadership challenges. We know that “big systems” make terrible decisions because of incentives, hurdles and compromises; therefore, why are we even considering outsourcing logic and reason to AI, as logic and reason drive decision-making?  

How do you know if anyone in your organisation is utilising any LLM to summarise facts and data to help with what should be presented to you?  Why do they do this? Because they are too busy as we already reduced headcount, the issues are complex, or there are other priorities that directly affect their KPI and bonus!

Based on the text so far, it is easy to think, let’s stop, but when the world is rapidly adopting AI technology, not being at the table is not an option. Part of your efficiency gains last year may have already come from outsourcing logic to a machine. This dilemma is one we all face, and the argument for and against outsourcing logic and reason to AI has several compelling dimensions:

Computational Advantages

  • AI can process vastly more information simultaneously
  • Lacks human cognitive biases
  • Can identify complex patterns humans might miss
  • Operates without emotional interference
  • Potentially more "objective" in pattern recognition

Potential Limitations

  • AI logic and reason are based on training data, which inherently contains biases - some good and much not so good.
  • Lacks true understanding of context and lived experience
  • Cannot generate genuine creativity or emotional intelligence
  • Risk of amplifying existing systemic biases encoded in training data
  • No genuine comprehension of ethical nuance

Philosophical Considerations

  • Does outsourcing reason to AI suggest a fundamental redefinition of intelligence?
  • Are we delegating reasoning or abdicating human cognitive responsibility?

  • AI "reasoning" is fundamentally pattern matching, which is not true reasoning and the logic is broken.
  • We would be substituting computational efficiency for “wisdom”
  • Who asks what is right and if we are doing the right thing?

Critical Risk Assessment

  • Do you/ your team believe that AI reasoning can be neutral or infallible?  
  • How do you unpack the sophisticated reflection of its training
  • Discus if it is potentially more dangerous because it appears more "objective."
  • How do you test for highly convincing but fundamentally flawed arguments?

The most responsible approach isn't wholesale outsourcing but collaborative reasoning—using AI as a powerful analytical tool while maintaining human ethical oversight, contextual understanding, and ultimate decision-making responsibility. However, this approach is inefficient and ineffective. While AI can calculate probabilities and identify patterns, humans decide what those patterns mean and what actions to take. However, humans can see patterns and connections that no AI can fathom, and machines are less influenced by incentives and politics to take the best action.

Reason isn't just about processing information—it's about understanding purpose, meaning, and ethical implications. These domains remain under threat but also offer the biggest opportunity.

The takeaway 

Embrace complexity, challenge your assumptions, and recognise that the most powerful insights often emerge from the spaces between our traditional categories—not from the categories themselves.

Would you like to explore how this new way of thinking might revolutionise your approach to communication, technology, or strategy? Then, start with questions.

Darragh Power

In my element helping you be in yours | Sketchnotes | Insight Principles | Changemaker

1w

Yes - 'challenge your assumptions, and recognise that the most powerful insights often emerge from the spaces between our traditional categories—not from the categories themselves' a dialogue not dependencies!

Aksinya Staar

🌍 Pᴏʟʏᴍᴀᴛʜ Mɪɴᴅsᴇᴛ Sᴛʀᴀᴛᴇɢɪsᴛ | Author | Sᴘᴇᴀᴋᴇʀ | Fᴜᴛᴜʀɪsᴛ | Board Advisor

1w

Brilliant insights! I particually love this: "We’re moving from a world of “this is true (logic)” to a world of “this is one possibility (reason).” And in that possibility lies an incredible opportunity for some because of uncertainty — but also a new risk at a scale where your brand can be lost in a day."

Phil Middleton

Providing a forum for directors from UK insurance to meet, benchmark and discuss technology and claims strategy.

1w

Thank you for these insights Tony Fish, recommended reading for attendees of the #DataJam

To view or add a comment, sign in

More articles by Tony Fish

Insights from the community

Others also viewed

Explore topics