I design decision architectures for human-AI systems—who decides what, when, and with what authority.

This work grows from 15+ years in service design, civic tech, and HCI research. Most system failures aren't interface problems—they're judgment routing problems. Wrong actor, wrong information, wrong authority structure.

Research areas:

• Preference architecture: Constraint design that reveals true intent (Stratified Preference Allocation)
• Trust infrastructure: Portable reputation and shared moderation (Glowrm)
• Agentic economics: The hidden costs of AI decision-making (Occupant Index)
• Disclosure systems: Making AI decisions legible for public accountability (Tardigrade, AI Statements)
• Delegation frameworks: When humans stay in the loop, when they don't (Judgment Routing)

The core insight: Traditional design assumes human decision-makers. AI doesn't just change the interface—it changes who decides. My work formalizes what designers used to do intuitively: structuring authority, information flow, and trust.

• Assistant Professor of Practice in Urban Technology, Taubman College, University of Michigan
• Board Member, State Capacity AI
• State Tennis Co-Chair, OACA (2023–)
• Founding Director, PDX Digital Corps (2025)
• Board President, AIGA Portland (2023–25)
• Co-Founder, Portland Design Month (2024)

Stratified Preference Allocation (SPA) — Matching architecture that replaces unlimited swipes with scarce, tiered signals. Forces honest preference revelation, reduces spam. Works across dating, hiring, networking—anywhere matching happens.

Reference implementations: HeyPBJ (dating) · Leafroll (professional networking)

Glowrm — Trust infrastructure for ATProto: portable reputation, shared moderation, SPA-based resource allocation

KizuKizu — Relational profiling around connection archetypes. How you connect, not just who you are.

Judgment Routing — Prototypes for delegated authority in agentic systems. When should a machine decide? When should a human?

Occupant Index — AI cost intelligence tracking the "Judgment Premium" in reasoning model pricing

Reasoning-class models (o1, DeepSeek-R1) have different economics than commodity inference. The Occupant Index tracks what aggregate "AI cost deflation" narratives miss: organizations building on AI judgment are taking an unhedged position on pricing that isn't deflating like they expect.

Portland Digital Corps — Civic tech sprint. Six Oregon organizations, 100+ participants, shipped projects.
GitHub · Final Report

Design For The Public 2024 — Two-day conference on public interest design and service delivery

Tardigrade �� Open disclosure patterns for AI systems: badges, contestability, human oversight indicators

AI Statements — Standardized AI usage disclosures

Consequence Design — Framework for post-deployment harms, institutional erosion, and the DIRE methodology

Oregon HS Tennis Rankings — Power Index rankings, playoff simulator, league analysis

A complete analytical system: adjusted win percentage (APR), strength of schedule (SOS), flight-weighted scoring (FWS). Fair, transparent team rankings. Built because ranking systems are decision architectures—they encode what we value and who gets opportunities.

Bluesky · Literal · Record Club · Last.fm · Mastodon · Mixcloud