Additive Bias: Why We Add Instead of Subtract

You are given a LEGO structure — an arch — and told to make it stable enough to support a heavy brick. The obvious problem is that one side of the arch has a single pillar that is shorter than the other, causing the arch to be lopsided. You can fix this by adding a pillar on the short side, or by removing the longer pillar on the other side. Either works. But nearly everyone adds. In experiment after experiment, people overwhelmingly choose to add elements to fix a problem rather than remove them — even when subtraction is the simpler, cheaper, and more elegant solution. This is additive bias, and it shapes everything from LEGO structures to tax codes to software products.

The 2021 Nature Study

The systematic study of additive bias was published in April 2021 in Nature by Gabrielle Adams, Benjamin Converse, Andrew Hales, and Leidy Klotz. The paper, titled "People Systematically Overlook Subtractive Changes," presented a series of experiments that documented the bias across remarkably diverse contexts.

In the LEGO experiment, participants were given the unstable arch and told they could add or remove bricks to make it stable; each added brick cost them ten cents while each removed brick was free. Despite the financial incentive to remove, 60% of participants added bricks. When the instruction made subtractive solutions more cognitively accessible — by reminding participants that they could remove bricks — the proportion choosing subtraction rose, but addition still dominated.

In a second experiment, participants were given a miniature golf hole and asked to improve it. Again, the modal response was to add obstacles, features, or decorations rather than to simplify by removing existing elements. In a third study, participants read a travel itinerary that was overly packed, with too many activities and insufficient rest time, and were asked to improve it. Most added activities to the gaps rather than removing activities to create space.

In a university context, Adams and colleagues analysed nearly 1,000 suggestions submitted to the president of a large university for improving the institution. Fewer than 10% of suggestions proposed removing something. The vast majority proposed adding: new programmes, new facilities, new requirements, new initiatives. People default to addition when they think about improvement, even when the existing system is already complicated and the most effective improvement might be simplification.

Why We Default to Addition

Adams and colleagues proposed several interacting explanations for why addition dominates in problem-solving and improvement contexts.

Cognitive Availability

Additive changes are more cognitively available than subtractive ones. When we think about improving something, the mental action of adding is more naturally activated than the mental action of removing. This may be partly because adding is the dominant mode of action in many everyday contexts: we add food to cook a meal, add words to write a document, add strokes to a painting. The cognitive operation of generating new elements is practised and fluent; the operation of questioning and removing existing elements is less rehearsed. The result is that additive options are generated first, evaluated first, and — absent deliberate search for subtractive alternatives — often chosen by default.

The Completeness Heuristic

Improvements, in most people's intuitive model, involve making something more complete, more developed, more filled-in. An empty space in an itinerary feels like a problem to be solved rather than breathing room to be preserved. A simple policy feels like an underdeveloped one rather than an elegant one. Subtraction can feel like regression, like giving up, like making something less rather than better. The heuristic "better = more complete" biases attention and evaluation toward additive solutions before subtractive ones are seriously considered.

Visibility and Credit

Additions are visible. When you add a rule, a feature, or a process step, something new exists that didn't exist before; its presence is evidence that work was done. Subtractive improvements are invisible: the removed rule, the deleted feature, the eliminated step leave no trace. In organisational contexts, where people are evaluated by their contributions, subtractive changes may be undervalued precisely because they leave no positive artefact. A manager who simplifies a process by removing five unnecessary steps may receive less credit than a manager who introduces a new system — even if the simplification was more valuable.

Loss Aversion

Removing an existing element can trigger loss aversion, particularly when the element was someone's contribution, has been in place for a long time, or is associated with a visible past investment. The sunk cost fallacy compounds this: once something has been added, removing it requires overcoming not just the psychological difficulty of loss but the additional reluctance to "waste" the effort that went into adding it. This makes organisations and systems particularly prone to accumulation over time — additions are easy to make and hard to reverse, while subtractions are difficult to make even when clearly warranted.

Feature Creep in Software

Software development provides one of the most visible and economically costly manifestations of additive bias. Feature creep — the gradual accumulation of features, options, and settings beyond a product's core purpose — is a nearly universal phenomenon in commercial software. The mechanism is almost always additive: users request new capabilities, stakeholders propose enhancements, product managers want competitive parity with rivals. Each addition makes sense in isolation. The cumulative result is software that is slower, harder to learn, more expensive to maintain, and less useful for the core tasks it was originally designed to perform.

The history of word processors illustrates the pattern vividly. Microsoft Word began as a simple writing tool and has accumulated, over four decades, hundreds of features that the large majority of users never touch. The ribbon interface introduced in Word 2007 was partly an attempt to make the accumulated features more navigable — an additive solution to the problem created by prior addition. The more radical subtractive solution — removing rarely used features and simplifying the interface — was the approach taken by new entrants like iA Writer and Hemingway Editor, whose market success suggests significant unmet demand for subtraction.

In software engineering, the concept of "technical debt" captures a related phenomenon: the accumulation of code complexity, workarounds, and legacy structures that results from years of additive patching rather than subtractive refactoring. Paying down technical debt — removing and simplifying — is the maintenance discipline that most development teams struggle to prioritise against the constant pull of new feature development.

Organisational Bureaucracy

Organisations accumulate rules, processes, approval layers, and reporting requirements through a consistently additive mechanism. Each addition was introduced in response to a real problem: the rule prevents the recurrence of a past mistake, the approval layer catches errors, the reporting requirement provides oversight. But organisations rarely have a corresponding subtractive mechanism: a systematic process for identifying rules that are no longer needed, processes that have been superseded, or oversight layers that add cost without value.

The result, familiar to anyone who has worked in a large organisation, is an ever-growing body of policy that nobody fully understands and that frequently contains internal contradictions from different additive epochs. Studies of organisational complexity have found that the burden of compliance with internal rules can become a significant drag on productivity — not because any single rule is unreasonable, but because their cumulative weight exceeds the organisation's capacity to follow them all coherently.

Amazon's "two-pizza team" rule — keep teams small enough to be fed by two pizzas — is one structural attempt to limit additive growth in organisational size. Jeff Bezos's famous "no PowerPoint" rule was a subtractive intervention in meeting culture: removing a standard communication medium to force more direct thinking. These examples are notable partly because they are unusual: organisations that deliberately subtract are exceptions in a landscape of habitual addition.

Law and the Ratchet Effect

Legal systems exhibit perhaps the most extreme form of additive bias, for structural reasons that compound the cognitive ones. Legislatures can pass new laws easily; repealing existing laws requires overcoming the political resistance of everyone who benefits from them. The result is a secular accumulation of legal text that rarely contracts. In the United States, the tax code exceeded 70,000 pages by the 2010s; the Code of Federal Regulations has grown almost continuously since its creation in 1938. Tax "reform" proposals that claim to simplify the code have, historically, added more provisions than they removed.

This ratchet effect in law is partly structural (repeal is harder than enactment) and partly cognitive (additions are visible improvements while removals look like losses of protection). The consequence is regulatory systems of extraordinary complexity that impose compliance costs on businesses and individuals that often exceed the value of the regulations themselves — a problem that cannot be solved by further addition.

Designing for Subtraction

Adams and colleagues found that additive bias can be reduced, though not eliminated, by cognitive interventions that increase the accessibility of subtractive options. Explicitly prompting people to consider what could be removed — before they generate improvement ideas — significantly increases the proportion of subtractive solutions. Organisations can institutionalise this by including "what should we stop doing?" as a standing agenda item alongside "what should we start doing?"

The concept of "zero-based budgeting" — requiring every budget line to be justified from scratch each cycle rather than treating existing expenditure as a default — is a formal attempt to counteract additive bias in resource allocation. "Sunset clauses" in legislation — provisions that automatically expire unless explicitly renewed — are a subtractive mechanism built into law that forces periodic reconsideration of whether rules should continue to exist.

The deeper challenge is cultural: in most organisational and social contexts, addition is associated with progress and subtraction with retreat. Building cultures that celebrate elegant simplification as much as innovative addition requires a deliberate reframing of what "improvement" means. The LEGO arch is perfectly stable with one pillar removed. Sometimes the best solution is the one that leaves less behind.

Sources & Further Reading

• Adams, G. S., Converse, B. A., Hales, A. H., & Klotz, L. E. "People Systematically Overlook Subtractive Changes." Nature 592 (2021): 258–261.
• Klotz, L. Subtract: The Untapped Science of Less. New York: Flatiron Books, 2021.
• Kahneman, D. Thinking, Fast and Slow. New York: Farrar, Straus and Giroux, 2011. (Chapter on loss aversion.)
• Schwartz, B. The Paradox of Choice: Why More Is Less. New York: HarperCollins, 2004.
• Spolsky, J. "Things You Should Never Do, Part I." Joel on Software (2000). (On software rewrites and additive impulses.)
• Wikipedia: Feature creep