Craig Roberts

Bringing Social Circles To Dating

2025
Product
  • Consumer
  • Social
Platform
  • iOS/Android
Scope
  • Product Strategy
  • Experience Design
  • Research Strategy
  • Org Leadership
What I did

I owned the problem that dating felt lonely and AI didn’t understand people deeply enough, shifted the experience from solo to social, and built a system that increased real dates and established a new, fully funded product pillar.

Role

Staff Product Designer

Release Date

August 2025

Runtime

Approx. 3–4 months

Cast
  • Hinge Labs
  • Hinge Exec Team

Hinge Social Team

User Problem

Dating makes people feel lonely and unseen because apps are built around individual experiences and shallow signals that fail to understand or reflect who they really are.

Org Problem

Hinge’s systems optimize thin signals like photos and prompts without access to real human context.

Business Problem

AI risked reinforcing shallow optimization, limiting its ability to improve dates, retention, and long-term differentiation in a crowded dating market.

The Finale

Product Impact

Increase in Hinge’s north star metric, dates

System Impact

Social context becomes AI input

Org Impact

Social becomes product pillar

Friend feedback increased the number of dates on Hinge. People updated their profiles based on input from others, which led to better matches and more real-world dates.

Testimonials added a layer of credibility by showing what others said about you, addressing the need for proof, not just self-description. This made profiles both more effective and more trustworthy, while also driving growth by bringing new users into Hinge through participation.

A social endorsements concept is presented as a core product strategy pillar at the Waterfront special event space at Chelsea Piers, in New York, NY. The presentation reflects demonstrated evidence that friends surface critical human signals AI cannot capture alone, leading social support to become a fully funded 2026 strategy pillar with a dedicated team.

7 Months Earlier

Team

Labs Team (Lead PD, Staff Data Engineer, Lead UXR)

Location

New York, NY Hinge HQ, Labs Office Enclave

Stakes

Whether AI could address loneliness, not just efficiency

By this point, I had built live AI prototypes that let teams test ideas directly, shifting research from slides to working experiences and earning leadership trust in the output.

As we used these with real data and real daters, I identified the core gap. The issue wasn’t AI capability. We were missing the human signals needed to understand people, which limited how much the system could actually help.

An employee works quietly in the Hinge headquarters in New York on a weekday afternoon. The space becomes a setting for early-stage AI exploration, where ideas are developed in isolation before being tested publicly and translated into product decisions.

The Conflict

Tension

Dating designed as a solo activity

Constraint

No authentic signal in the system

Risk

AI optimizing without understanding

Dating products are built as solo experiences, but people don’t date alone. They send screenshots to friends, ask for advice, and even borrow photos before making decisions. Hinge had no way to capture this input.

As a result, profiles reduced people to a few prompts, and the system had no access to how they were actually perceived by others. AI could only learn from what individuals said about themselves, not how they showed up in real relationships.

The question became, could bringing trusted people into the product give AI the context it needed to better understand and help people?

An article from The Atlantic, “The Dating-App Diversity Paradox,” is read and analyzed in a research setting at Hinge, in New York, NY. The reading reflects ongoing research into how people actually experience online dating, highlighting a gap between products designed for individual use and dating as it often happens socially through shared profiles, screenshots, and group conversations.
A Hinge employee observes a live user research session as real daters use an early prototype with their own profile data in a research room at the company’s offices, in New York, NY. The session reflects the first time Hinge placed unfinished features directly into users’ hands, allowing teams to discuss impact through how people experienced changes to their real profiles rather than abstract strategy alone.

The Climax

Action

Make loneliness testable

Mechanism

Social input + AI synthesis

Shift

Individual → relational understanding

I introduced friend input into the product so AI could learn from how people are actually seen by others, not just how they describe themselves.

To make this testable, I designed social features and built prototypes that let multiple people interact with the same profile in real time, so teams could observe how friends influenced decisions and behavior.

This shifted the system from relying on self-presentation to learning from real relationships, giving AI access to signals profiles consistently missed.

A prototype profile built on real dater data runs during the AI Vision Sprint at Hinge headquarters in New York. The internal AI API, developed in partnership with engineering, allows concepts to be tested in live product conditions, marking a shift from speculative decks to experiences the company could evaluate and trust.
Daters respond to friend-focused profile features during an in-person friends research session at Hinge offices, in New York, NY. The session reflects a first-of-its-kind study at the company that captures positive reactions to friends actively helping one another build more authentic and effective dating profiles.

Details

Genre

Product Pillar Formation

Audience
  • Design Leaders
  • Product Leaders
  • Engineering Leaders
  • Founders
  • Companies looking to improve their AI systems and experiences
Tagline

“AI didn’t fail. It was never given the right signal.”

Team Budget(s)

Est. $450–600K

Est. $8–15M in strategic value