Addendum: Refining Cognitive Diversity – The ‘Who’ Matters as Much as the ‘What’
...and Who Sees Why Newcastle United Struggle Against a Low Block
In the spirit of “seeing further” through planned cognitive diversity - the core idea from the main piece - let’s extend the discussion here, in what is admittedly a ‘half post’ more than a post in itself…
I’ve long believed that in pharma, simply gathering intelligent people doesn’t deliver the edge we need. What matters is asymmetric learning: the kind that can boost creativity by around 20% and reduce risks by 30%. But that only happens when the diversity is deliberate - carefully selecting the right “who” to address specific perceptual gaps, then assigning differentiated tasks that draw on their particular strengths. Without that intentional design, even a highly interdisciplinary group can fall into the symmetric traps we’ve discussed: echo chambers, or insights that are essentially commoditized because they’re drawn from the same shared data pools.
The ‘Who’ Factor: Targeted Expertise as Your Asymmetric Edge
Cognitive diversity isn’t about diversity for its own sake, nor is it a matter of adding an outsider simply to introduce novelty. It’s about identifying and recruiting the precise perspectives that expose blind spots in your current R&D approach. The Manhattan Project succeeded not because it assembled a collection of brilliant minds at random, but because it brought together physicists, engineers, and chemists whose complementary expertise created genuine consilience.
In pharma, the principle is the same. If your discovery team is dominated by biologists focused on molecular targets, introduce a behavioral economist with deep experience in patient adherence modeling - they can reveal why structurally similar compounds failed in the clinic due to compliance factors that laboratory data never captured. Or, as in the cross-disciplinary pods at Roche I referenced earlier, include a data ethicist familiar with the FDA’s 2025 AI guidance (the draft on AI supporting regulatory decision-making for drugs and biologics). That person can identify dataset biases at an early stage, helping to avert later-stage attrition.
This targeted selection is what enables planned serendipity. A former journalist on the team, for example, might surface proprietary market intelligence from unconventional sources during an informal conversation, opening a strategic angle that competitors overlook entirely.
The soccer analogy resonates strongly with me. During a recent Newcastle United halftime analysis, the ex-professionals on the panel - chosen not for superficial variety but for their accumulated expertise - dissected midfield dynamics and defensive vulnerabilities in ways that would escape even a dedicated viewer over many years. (I had been watching the game through my fingers, and with a beer or two, but it wouldn’t have helped…) Their perceptual and cognitive edge, built through thousands of hours on the field, allows them to deconstruct a goal’s buildup by focusing on subtle cues: body orientation, off-ball movement, shoulder positioning. Research on elite performers shows they anticipate plays 20–50% more accurately than novices, thanks to tracking patterns that prioritize predictive signals over the ball itself. A casual fan’s halftime take is typically surface-level - ”they’re off today” - while a professional exposes the underlying tactical mismatches, such as how a defender’s hesitation created exploitable space - and these are the same choices made on the field that can be exploited by a determined competitor. The lesson for pharma is straightforward: Audit not only for diverse backgrounds, but for demonstrated cognitive depth. Tools like skill-mapping or perceptual profiling can ensure your selections deliver genuine asymmetric insight rather than incremental commentary.
Task Differentiation: Assigning Roles to Elicit Unique Views
Having the right people is necessary but insufficient. Generic assignments dilute their potential. The structured debates and integration emphasized in my main piece are valuable, but the real leverage comes from differentiating tasks to surface distinct perspectives, then synthesizing them through consilience.
Consider a compact biotech team modeled on Lilly’s approach: Assign the chemist to molecular simulations, the AI specialist (drawing from collaborations like NVIDIA-Merck) to pattern extraction in large-scale datasets such as the KERMT model (pretrained on over 11 million molecules for early ADMET prediction), and the clinician to real-world patient subpopulation scenarios that data-driven analyses might undervalue. This mirrors the risk-based accountability principles in EMA-FDA guidance on AI governance, where roles are allocated deliberately - an ethicist evaluates transparency, a modeler assesses credibility.
On the pitch, halftime analysts don’t offer uniform narration. An ex-defender focuses on positional errors and spatial exploitation, while a former midfielder unpacks passing structures and decision-making flows. Studies confirm that experts scan the field more dynamically, detecting anticipatory cues (weight distribution, preparatory movements) that lead to decisions 30% faster than in novices. Casual observers attribute outcomes to chance; professionals trace systemic causes across the half. In pharma, adapt similar differentiation: Extend frameworks like RACI by designating “Perceiver” roles to uncover latent risks and “Synthesizer” roles to integrate findings. This structured approach amplifies planned serendipity, converting diverse inputs into proprietary advantages - such as identifying a novel biomarker ahead of the field.
Pitfalls, Measurement, and Making It Stick
The risks are real. Misjudging the “who” - over-relying on senior voices and marginalizing junior perspectives, for instance - or assigning mismatched tasks can reinforce biases, undermine psychological safety, and impair coordination. Move beyond output metrics alone. Track insight asymmetry directly: the number of novel hypotheses generated, or whether the team identified critical issues (a manufacturing vulnerability, say) significantly earlier than in prior projects.
Implementation begins with a cognitive diversity audit, followed by piloted workshops with differentiated tasks. In AI-intensive R&D environments, integrate the FDA’s 7-step credibility assessment framework to ensure diverse contributions support robust, bias-mitigated decisions.
Ultimately, this extension reinforces a central point: Cognitive diversity in pharma is not an organic outcome; it is engineered. By curating the “who” with precision and designing the “how” through thoughtful task allocation, teams can achieve true asymmetry - gaining the vantage point to see further and advance faster in an industry where symmetric approaches leave most players trailing.
(This asymmetric learning does seem to apply to Newcastle United’s coaches, but in the wrong direction 😣)