Why researchers needed an ontology for dark patterns

The field has a vocabulary problem

Ask three specialists to name the same troublesome interface and you may receive three answers. A UX researcher might describe “false hierarchy”. A lawyer might focus on a misleading omission. A privacy specialist may call the same screen “left in the dark”. Each description can illuminate part of the experience, but the team can lose the underlying evidence while debating labels.

The ontology paper tackles that translation problem. Its authors brought together ten academic and regulatory taxonomies and organised their concepts into three levels. That structure is less about inventing new names than about showing how existing names relate.

Three levels, three different questions

A low-level pattern can describe what is visible: a preselected checkbox, an obscured link or a countdown. A meso-level concept can describe the strategy, such as obstructing a route or interfering with comparison. A high-level concept can describe how autonomy or decision-making is undermined.

These levels should not be collapsed. A team can observe a preselection without knowing its effect on every user. It can recognise an obstructed route without yet deciding whether a legal threshold is met. The ontology helps keep those statements connected but distinct.

What this portal borrows from that approach

The glossary stores synonyms as aliases rather than publishing duplicate pages. Each canonical record then connects the name to a family, journey stages, examples, a fairer alternative and cautious legal context. That design makes search easier without pretending that terminology is perfectly settled.

The seven-family structure is deliberately a usable learning layer, not a claim that seven families are the only valid academic classification. A person arriving with an EDPB, OECD or research term should still reach the relevant mechanism.

Where the legal work begins

An ontology can tell a legal team that several labels appear to describe related choice architecture. It cannot identify the applicable provider role, material scope, likely effect or evidence required by a provision. Those questions belong in the legal map and the complete journey record.

The most productive meeting therefore starts with a sentence that names its level: “We observed this control,” “we think it implements this strategy,” or “our research suggests this effect.” That small discipline stops a label from doing more work than the evidence supports.

A living map, not a final dictionary

The authors present the ontology as infrastructure for future knowledge. New domains, devices and agent-mediated interfaces will create variants that older taxonomies did not anticipate. A useful portal should therefore maintain stable concepts and URLs while allowing definitions, aliases and examples to improve as evidence develops.

Source check on 14 September 2026

The publisher version in Utrecht University’s repository confirms CHI 2024 publication. Its abstract says 64 types, while section 3.2.1 and the conclusion say 65, comprising five high-level, 25 meso-level and 35 low-level patterns. This explainer follows the explicit breakdown and records the discrepancy. The earlier arXiv abstract also says 65; this is not evidence of a new 2026 taxonomy.

Questions for a journey review

1. Are two teams discussing the same interface mechanism under different labels, or different mechanisms under one broad label?
2. Is the review describing a visible control, an underlying strategy or a claimed consumer effect, and is that level stated clearly?
3. Can synonyms remain searchable aliases while one stable canonical concept preserves evidence and internal links?