OceanCanada

Problem framing

OceanCanada is a national research partnership producing peer-reviewed work on fisheries, ocean ecosystems, climate change, and coastal communities. Despite strong research output, the organization faced a familiar challenge in high-accountability, data-heavy environments:

Critical information existed, but the system did not consistently enable correct interpretation, publication, or reuse by non-experts, including the public, students, policymakers, and organizational staff.

This project approached knowledge translation as a product and systems problem, not a content problem.

Within OceanCanada, knowledge translation relied heavily on individual judgment, manual effort, and tacit expertise held by a small number of people. This created variability, bottlenecks, and institutional risk. The outcome was a scalable, governance-aware platform concept that enabled public users to understand complex scientific research without domain expertise, researchers to preserve scientific integrity, non-design, non-UX admins to curate, translate, and publish content independently, and the organization to operate within copyright, credibility, and institutional constraints.

Design problem

This work extended beyond end-user usability into organizational enablement. Key characteristics of the problem included:

• Multiple user types with different authority and accountability

• Information where misunderstanding had real reputational and societal consequences

• Publishing workflows owned by non-design, non-research staff

• A need for consistency and trust over time, not one-off engagement gains.

Design question

How do we design a system that enables correct decisions at scale by people who are not designers, researchers, or technologists?

Constraints

This project was shaped by non-negotiable constraints common to high-accountability information systems. These constraints defined the design space and informed every structural decision.

Technical constraints

• Implementation limited to UBC’s WordPress-based CMS

• No custom backend services or automated ingestion pipelines

• Design solutions had to align with existing CMS patterns and permissions.

Operational constraints

• Administrators manually populate publication pages

• Content sourced from peer-reviewed articles and academic books

• Admins have no formal training in UX, data visualization, or knowledge translation.

Interpretation and data constraints

• Visualizations generated from a controlled, preset taxonomy of fisheries and ocean-science keywords

• Keyword lists designed to preserve semantic accuracy, prevent speculative interpretation, and maintain consistency across publications.

Design objective

Our goal was to embed knowledge translation into the platform, so that correct interpretation and publication decisions emerge from the workflow.

Note: Success was not defined by visual polish, but by reduced ambiguity for admins, consistent outputs across time and contributors, lower cognitive and operational load, and preservation of scientific credibility.

Research and synthesis

I led synthesis across a design study process combining surveys with students and general audiences, exploratory sketch exercises to understand sense-making, workshops with OceanCanada researchers, and usability testing across expertise levels.

Key insights

1. More information reduced engagement when context was missing: Users needed relevance before depth, orientation before detail, and multiple, intentional entry points.

Design requirement: The system must stage understanding, not simply expose data.

2. Knowledge translation is a workflow and governance problem: Researchers identified five recurring areas of action:

1. Communicating research outside academia

2. Simplifying without distorting meaning

3. Public outreach via alternative media

4. Information verification across channels

5. Coordinated knowledge dissemination.

These insights translated into design requirements focused on information practices, not just interfaces, including navigating multiple levels of complexity, enabling semantic consistency across media, validating information relevance, and supporting timely, coordinated publishing.

Design requirement:
 The system must enable good knowledge-translation behaviour, rather than depend on individual expertise.

Design Strategy: Decision-Making & Enablement

1. Admin enablement as a design problem

Admins were treated as decision-makers, not content uploaders. The system enabled them to:

• Identify which elements of a publication were relevant to non-expert audiences

• Associate content with predefined topics and regions

• Publish at appropriate levels of complexity

• Maintain semantic consistency without UX or research training.

2. Visualization supporting interpretation

Visualizations functioned as interpretive aids, not analytical endpoints, reducing misinterpretation while preserving meaning. Design choices included:

• Controlled keyword taxonomies

• Explicit legends and instructional cues

• Visual hierarchy emphasizing relationships over metrics.

Aligning UI language with institutional realities

We designed several screens in low and high-fidelity mock-ups to show users for feedback.

Low fidelity sketches

Article page wireframe

Outcomes

Organizational outcomes

• Reduced reliance on specialized expertise

• Increased consistency and reliability in published content

• Improved alignment between researchers, admins, and public audiences.

Platform outcomes

• Shift from static repository to interpretive system

• Knowledge translation embedded into everyday workflows

• A model that scales through structure rather than training.

Validation and iteration

Usability testing showed that visual summaries lowered entry barriers, filters significantly improved discovery, and users required explicit guidance to interpret charts correctly. Iterations focused on: clarifying semantics, improving regional context, and distinguishing open-access vs. restricted content.

Future considerations

While automation was limited at the time of the project, the platform was intentionally designed to be extensible. Its structured content model, controlled taxonomies, and decision-oriented workflows create a strong foundation for responsible system-level augmentation.

Today, a combination of AI and deeper system integration could further reduce cognitive load, improve consistency, and strengthen governance, while preserving human accountability. A few orientations include:

1. Integrated metadata extraction from reference managers

Integration with reference management tools (e.g., Zotero, Mendeley) could enable, automatic extraction of article metadata (authors, publication date, journal, DOI), reduction of manual entry errors, and faster, more consistent population of publication pages. In doing so, the system could enable admins to spend less time transcribing information and more time validating relevance and context.

2. Built-in author permissions and validation workflows

The system could support, structured permission requests to article authors prior to publication, inline validation or feedback loops for summaries and interpretations, and clear visibility into approval status within the CMS. System Governance and credibility are enforced through workflow, not informal coordination.

3. System-generated keywords (human-reviewed)

Building on the existing controlled taxonomy, the platform could suggest keywords derived from article content, constrain suggestions to approved vocabularies, and require explicit admin confirmation before publication. System consistency would be improved without removing editorial judgment.

4. Expanded chart types and controlled customization

Additional visualization types and limited customization options could be introduced, incorporating alternative chart forms selected based on content type, guard-railed configuration (e.g., scale, grouping, emphasis), and presets designed to prevent misleading representations. Admins could gain expressive flexibility while the system prevents risky interpretation.

5. Geographic relevance and impact visualization

The platform could more explicitly surface geographic context by visualizing where studies were conducted, indicating regions or communities impacted by the research, and supporting multiple geographic layers (local, regional, global). In doing so, users would gain situational understanding without requiring deep domain knowledge.

6. AI-generated lay-language summaries

Building on the original knowledge-translation goals, AI tools could draft plain-language summaries from scientific text, highlight assumptions, limitations, and uncertainties, and flag areas requiring human review before publication. The system would support admins in translation tasks without delegating authority to automation.

In all cases, AI features act as decision support, not autonomous publishing mechanisms. Human review, approval, and responsibility would remain explicit and traceable.

Takeaway

This project reframes a public-facing research platform as a high-accountability decision-making system. The design prioritizes operational consistency, governance, and decision quality, enabling non-expert operators to publish and maintain complex information responsibly under technical and institutional constraints.