← Latest papers
💻 computer science

Epistemic Trust as a Mechanism for Ethics Integration: Failure Modes and Design Principles from 70 Moral Imagination Workshops

Drawing on qualitative analysis of 70 moral imagination workshops, this paper proposes "epistemic trust" as a critical mechanism for successful ethics integration, identifying five key dimensions, 23 failure modes, and nine design principles to explain why some responsible innovation initiatives succeed while others fail to engage practitioners.

Original authors: Benjamin Lange, Geoff Keeling, Kyle Pedersen, Carmen Heringer, Susan B. Rubin, Ben Zevenbergen, Amanda McCroskery

Published 2026-04-15
📖 6 min read🧠 Deep dive

Original authors: Benjamin Lange, Geoff Keeling, Kyle Pedersen, Carmen Heringer, Susan B. Rubin, Ben Zevenbergen, Amanda McCroskery

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Why Ethics Workshops Often Fail

Imagine you are a team of expert chefs building a new, complex robot that can cook meals. You are brilliant at coding and engineering. Suddenly, a group of "Ethics Chefs" walks in and says, "Hey, before you finish this robot, let's talk about whether it's right to feed people this way."

In the past, the engineering team usually rolls their eyes. They think:

  • "This is boring philosophy, not real work."
  • "You don't understand how our robot works."
  • "You're just here to stop us from innovating."
  • "This feels like a lecture from a boss who doesn't know what we do."

The result? The engineers tune out, check their phones, and the workshop fails.

This paper, based on 70 real workshops with Google engineers, asks: Why do some of these ethics sessions work like magic, while others feel like a waste of time?

The authors propose a new concept called Epistemic Trust. Think of this not as "trusting a friend," but as "trusting the source of information." It's the feeling engineers get when they think: "Okay, these people know what they are talking about, they get our job, and what they are saying actually matters to us."

If you don't have this trust, the engineers' brains shut down. If you do, they lean in and start thinking deeply.


The 5 Ingredients for Trust (The "Secret Sauce")

The authors found that to build this trust, you need five specific ingredients. If you miss even one, the recipe fails.

  1. Relevance (The "So What?" Factor):

    • The Problem: If the ethics talk is too abstract (like talking about "justice" in a vacuum), engineers think, "How does this help me fix this bug?"
    • The Fix: The conversation must be about their specific robot, their specific code, and their specific problems. It needs to feel like a tool, not a textbook.
  2. Inclusivity (The "No Bosses Allowed" Rule):

    • The Problem: In many meetings, the senior engineer or the manager talks the most. The junior staff stays quiet.
    • The Fix: You need a safe space where the quietest person feels brave enough to say, "Wait, I think this is dangerous," without getting fired or mocked.
  3. Agency (The "Driver's Seat" Feeling):

    • The Problem: If engineers feel like they are just being lectured or told what to do, they get defensive. They feel like passengers, not drivers.
    • The Fix: They need to feel like they are solving the puzzle themselves. They need to come up with the answers, not just listen to the facilitator's answers.
  4. Authority (The "Credibility" Check):

    • The Problem: If the ethics facilitator doesn't understand the tech, the engineers will say, "You don't know what you're talking about." If the facilitator is too technical, they might miss the ethical point.
    • The Fix: The facilitator needs to speak both languages: "Tech" and "Ethics." They need to sound like a peer, not a tourist.
  5. Alignment (The "Same Team" Vibe):

    • The Problem: If engineers think ethics is an enemy trying to block their project, they will fight back.
    • The Fix: They need to see ethics as a partner helping them build a better product, not a police officer trying to shut them down.

The 23 Ways Things Go Wrong (The "Failure Modes")

The authors analyzed 70 workshops and found 23 specific ways these sessions can crash. Here are a few colorful examples:

  • The "External Imposition" Trap: The team feels like the workshop was forced on them by management. They put up a wall immediately.
  • The "Abstract Fog": The facilitator talks about "human dignity" but never connects it to the actual code. The engineers stare blankly.
  • The "Silent Room": The boss speaks for 20 minutes, and no one else says a word. The team thinks, "Well, the boss is happy, so I'll just go along with it."
  • The "Naive Expert": The facilitator suggests a technical solution that is outdated or impossible. The engineers lose all respect for them instantly.
  • The "Checklist" Feeling: The workshop feels like a rigid form to fill out rather than a real conversation.

The Solution: 9 Design Principles

To fix these crashes, the authors came up with 9 rules for running a successful workshop. Think of these as the "User Manual" for ethics facilitators.

  1. Customize the Menu: Don't use generic examples. Use the team's actual project. If they are building a self-driving car, talk about self-driving cars, not generic philosophy.
  2. No "Gotcha" Moments: Frame the session as "Let's explore this together" rather than "Here is what you did wrong."
  3. The "Bilingual" Guide: The person leading the session must understand both the code and the ethics. If they don't, bring in a co-leader who does.
  4. Shuffle the Seating: Actively make sure the quiet junior engineers get to speak before the loud senior managers.
  5. Let Them Drive: Don't give them the answers. Ask them questions and let them figure out the ethical path.
  6. Challenge Gently: It's okay to push them on their ideas, but do it like a friendly sparring partner, not an enemy.
  7. Clear Goals: Make sure everyone knows why they are there. If the team thinks it's a team-building exercise and the facilitator thinks it's a compliance audit, you have a problem.
  8. Make it Real: Don't just have a nice chat. Turn the ideas into actual to-do lists or project plans (like OKRs) that they have to do next week.
  9. Explain the "Why": If you use role-playing (pretending to be a user), explain why you are doing it so it doesn't feel silly or childish.

The Big Takeaway

The paper concludes that you cannot force engineers to care about ethics. You have to earn their trust.

If you treat them like smart, capable experts who are trying to build something great, and you show them how ethics helps them build better things, they will engage. If you treat them like rule-breakers who need to be lectured, they will tune out.

In short: Ethics isn't a rulebook you hand to engineers; it's a conversation you have with them. And for that conversation to happen, you have to build a bridge of trust first.

Drowning in papers in your field?

Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.

Try Digest →