Skip to content

boundary-objects.adoc

Latest commit

 

History

History
292 lines (204 loc) · 8.38 KB

boundary-objects.adoc

File metadata and controls

292 lines (204 loc) · 8.38 KB
Raw

Boundary Objects: Shared Maps with Multiple Routes

What Are Boundary Objects?

The term comes from sociologists Susan Leigh Star and James Griesemer (1989):

Boundary objects are objects which are both plastic enough to adapt to local needs and the constraints of the several parties employing them, yet robust enough to maintain a common identity across sites.

— Star & Griesemer

The Map Metaphor

Think of a map. A tourist, a civil engineer, and a military strategist all use the same map—but for completely different purposes:

User What They See What They Need

Tourist

"Where are the restaurants?"

Routes, landmarks, attractions

Civil Engineer

"Where are the utilities?"

Infrastructure, elevation, drainage

Military Strategist

"Where are the chokepoints?"

Terrain, cover, sightlines

Same map. Different uses. The map works as a boundary object because it’s:

  1. Robust enough to maintain a shared structure (streets are streets)

  2. Flexible enough to support different interpretations

Evidence as a Boundary Object

In investigative journalism, evidence serves multiple communities:

Community Their Relationship to Evidence

Activists

"This proves systemic injustice"

Policymakers

"This informs policy decisions"

Researchers

"This supports or refutes hypotheses"

Skeptics

"This needs more verification"

Affected Persons

"This explains what happened to me"

Journalists

"This is what we can publish"

They’re all looking at the same ONS inflation data—but asking different questions.

Why Coordination Without Consensus?

The key insight: you don’t need everyone to agree on what evidence means to work together productively.

TRADITIONAL APPROACH:
  "Let's agree on what this evidence means"
  → Conflict, deadlock, power struggles

BOUNDARY OBJECT APPROACH:
  "Let's agree on what the evidence IS"
  → Each community interprets for their context
  → Coordination happens through shared structure

Example: Climate Data

Climate scientists, oil executives, and environmental activists can all use IPCC data:

  • Scientists: "This confirms our models"

  • Executives: "This informs risk assessment"

  • Activists: "This demands urgent action"

They’ll never agree on what to do about climate change. But they can agree on the data—and that’s enough to coordinate.

How Bofig Implements Boundary Objects

1. Single Source of Truth

All communities access the same underlying evidence graph. There’s one ONS dataset, one academic study, one interview transcript.

┌─────────────────────────────────┐
│       EVIDENCE GRAPH            │
│   (One shared structure)        │
├─────────────────────────────────┤
│  Claims ←→ Evidence ←→ Claims   │
│       Relationships             │
│       PROMPT Scores             │
└─────────────────────────────────┘
         ↓         ↓         ↓
    Researcher  Policymaker  Activist
      VIEW        VIEW        VIEW

2. Navigation Paths as Interpretive Lenses

Different audiences access the same data through different paths:

Path Type What It Emphasizes

Researcher Path

Methodology → Data → Analysis → Counter-evidence

Policymaker Path

Authoritative sources → Recommendations → Trade-offs

Activist Path

Impact → Systemic causes → Call to action

Skeptic Path

Weak points → Counter-evidence → Open questions

3. PROMPT Scores as Boundary Metadata

PROMPT scores work as boundary objects themselves:

  • A researcher interprets "Replicability: 70" as "needs validation"

  • A policymaker interprets it as "good enough for decision-making"

  • A skeptic interprets it as "where’s the verification?"

Same number, different meanings—but the number provides shared ground.

Star & Griesemer’s Four Types

Star and Griesemer identified four types of boundary objects:

Type Definition In Bofig

Repositories

Ordered "piles" of objects indexed for different uses

The evidence database

Ideal Types

Abstract diagrams or templates

PROMPT scoring framework

Coincident Boundaries

Common objects with different internal contents

Navigation paths

Standardized Forms

Methods of communication across groups

GraphQL API, JSON-LD exports

Bofig uses all four:

  1. Repository: ArangoDB stores claims, evidence, relationships

  2. Ideal Type: PROMPT dimensions define what "good evidence" means (abstractly)

  3. Coincident Boundaries: Same evidence, different navigation orders

  4. Standardized Forms: GraphQL mutations, Zotero imports, Dublin Core metadata

The Translation Problem

Boundary objects don’t eliminate translation—they make it visible.

WITHOUT BOUNDARY OBJECTS:
  Group A: "Energy drove inflation"
  Group B: "No, food prices drove inflation"
  → Deadlock (talking past each other)

WITH BOUNDARY OBJECTS:
  Shared Evidence: ONS CPI data, energy price records, food price records
  Group A's Path: Focuses on energy → concludes energy drove inflation
  Group B's Path: Focuses on food → concludes food drove inflation
  → Both paths visible, disagreement localized

The disagreement isn’t about what the evidence is—it’s about which parts matter most. That’s a productive disagreement.

Plasticity vs. Robustness

Boundary objects balance two properties:

Property What It Means

Plasticity

Can be shaped to local needs (different navigation paths, different PROMPT weights)

Robustness

Maintains core structure (same claims, same evidence, same relationships)

Too plastic → everyone interprets differently, no shared ground Too robust → one interpretation dominates, no local adaptation

Bofig aims for the sweet spot: shared data, multiple interpretations.

Practical Benefits

For Journalists

  • Work with sources who have different agendas

  • Present evidence without forcing readers into one interpretation

  • Make the basis of claims explicit

For Researchers

  • Share data with policymakers without oversimplifying

  • Collaborate across disciplines with different standards

  • Make methodology accessible to non-experts

For Policymakers

  • Access evidence without wading through academic jargon

  • See what different stakeholders prioritize

  • Make decisions with visible trade-offs

For Affected Communities

  • See how their experiences connect to systemic patterns

  • Access evidence in accessible formats

  • Challenge expert interpretations on shared ground

Example: Multi-Stakeholder Investigation

Imagine an investigation into housing costs:

EVIDENCE:
  - ONS housing price index
  - Shelter report on homelessness
  - Landlord association lobbying docs
  - Tenant interviews
  - Academic study on rent controls

STAKEHOLDERS:
  - Tenant advocates → Want to show exploitation
  - Landlords → Want to show market constraints
  - Policymakers → Want to identify interventions
  - Researchers → Want to test economic theories

SHARED GROUND:
  All stakeholders use the same evidence graph
  Each has a navigation path suited to their needs
  Disagreements are localized to interpretation, not data

Related Concepts

References

  • Star, S.L. & Griesemer, J.R. (1989). "Institutional Ecology, 'Translations' and Boundary Objects: Amateurs and Professionals in Berkeley’s Museum of Vertebrate Zoology, 1907-39." Social Studies of Science, 19(3), 387-420.

  • Bowker, G.C. & Star, S.L. (1999). Sorting Things Out: Classification and Its Consequences. MIT Press.

Last Updated: 2025-01