The failure modes of principled innovation

Paralysis, ethics washing, moral licensing, groupthink, and poorly managed trade-offs. The organisations that fail at this rarely lack principles. They have too many, or announce them, or feel they have earned some slack, or agree with each other.

1. Paralysis#

Endless deliberation stalls the work entirely. Principles are meant to guide rather than immobilise, and a review process that cannot reach a decision has replaced one failure with another. The organisations most prone to it are the ones that took ethics seriously late and overcorrected, and the symptom is a committee with no mechanism for saying yes.

2. Ethics washing#

Adopting the language of values for image without the follow-through. It is the most visible failure and the most punished, because the gap between statement and behaviour is discoverable by anybody and the discovery accelerates the reputational damage well past what the original decision would have cost.

The structural test is whether anything the organisation says it will not do has ever cost it anything, which is what a cost ledger makes checkable.

3. Moral licensing#

People who consider themselves highly moral can become more prone to later missteps, feeling they have earned some slack. This is the one that surprises people, because it means a strong ethical self-image is not a safeguard and can operate as the opposite.

It is also the argument for making the practice structural rather than dispositional. A team that has just published a responsible AI policy is, on this account, in a slightly riskier state than it was the week before, and the remedy is the machinery in the Integrity Loop rather than a renewed commitment.

4. Groupthink#

A values-driven team becomes insular and resistant to outside critique, and the shared commitment that makes it effective is the same thing that makes challenge feel disloyal. The remedies are the ones at collective judgement: independent judgements collected before discussion, and somebody in the room whose job is to argue the other side, which is what the absent user does for the people affected.

5. Poorly managed trade-offs#

Speed against thoroughness, safety against access, privacy against transparency. These cause failure in both directions and they are the most frequent of the five, because every real decision contains one.

The failure is rarely choosing wrongly. It is choosing without recording which way it went or why, so that the trade-off cannot be revisited when conditions change and nobody can say later whether it was considered at all. That is the same absence a decision journal addresses, applied to the question of principle.

What the five have in common#

None of them is an absence of values. Each is a lapse in practice: missing follow-through, missing perspective, or missing balance. That is why the remedies in this body of work are procedural rather than exhortative, and why a page on ethics that ends by recommending that people care more has diagnosed nothing.

What this does not establish#

The five are named from practice rather than measured, and nothing here counts how often each occurs. Moral licensing has an experimental literature behind it, and the other four are patterns described from observation. No study has tested whether the remedies attached to each one reduce its frequency.

Source#

Rahim Hirji, SuperSkills: The Seven Human Skills for the Age of AI, Kogan Page, 2026, chapter six.

Related SuperSkills research#

• Principled Innovation
• The Principle Test
• The Integrity Loop
• AI governance versus AI leadership