The myth of “human error” (and what it’s really hiding)

Alice loses her phone. Tied to that phone was her private recovery phrase, the only way to get back her cryptocurrency funds. There’s no password reset. No customer support line to call. The funds are just gone.

Her story isn’t rare. Phones are lost or stolen all the time, and each loss can mean losing the cryptographic keys tied to it, and with them, the funds.

Those keys usually take the form of a private recovery phrase (also called a seed phrase or mnemonic phrase), something most users have never encountered anywhere else in the products they use. Non-custodial wallets, the ones that let you control your own phrase, put the full burden of safe storage on you. Simply put, they’re not user-friendly. The alternative? Custodial wallets, where you trust another organization to hold your private recovery phrase for you.

As people designing products, we have a responsibility to ensure our products are both usable and secure. Our CEO, Colin Gounden, often emphasizes “no compromise.” Users shouldn’t have to choose between security and usability.

Alice not saving her private recovery phrase might be attributed to “human error.” But this isn’t human error. It’s a moment, or a series of moments, where the system failed to support the user.

The efforts to prevent human error are not new. The widely-cited Professor of Psychology James Reason, in his book Human Error, explains the different types of errors: mistakes (broken down into knowledge-based mistakes and rule-based mistakes), slips, and lapses.

Reason advises readers to ask: What can the system do to support the user? How can we prevent the negative thing from happening?

Here’s what each looks like in practice:

Knowledge-based mistakes happen when how a user thinks a system works (their mental model of the system) isn’t aligned with how it actually works. Many people don’t know to change their router settings’ default username and password (often shipped with defaults like username=admin and password=password) because most routers don’t guide users through the setup process. Users who don’t know a setting exists will never go looking for it.

Rule-based mistakes happen when something Alice has learned is not helpful in the particular situation she’s in. For example, in the past, Alice has shared a data visualization with Bob and Charlie but they were never able to see the underlying data (sensitive data that Bob and Charlie should not have access to). However, the new system Alice is using works a bit differently. So the next time she shares a visualization, Bob and Charlie can view the raw data, too.

Unlike a knowledge-based mistake, the rule Alice is following (“I’m only sharing what I made, not the data behind it”) usually serves her well. Just not here.

Slips are when Alice means to do one thing and instead does another. Alice intends to share a file containing confidential data with Bob Smith on her team, but sends it to Bob Smith the vendor instead.

Lapses are usually what we mean when we say “they forgot.” Alice knows she needs to do something and simply doesn’t. The more steps involved and the more pressure she’s under, the likelier the lapse. For example, Alice the IT team member usually checks the box to enforce two-factor authentication on all new users she sets up. Except, today she was interrupted during setup and misses the check box. Alice knows she needs to complete the step, she simply forgets.

Rule-based mistakes, slips, and lapses are hard, but they have common solutions: step-by-step guides, clearer labels, and confirmation dialogs. Knowledge-based mistakes are different: the user doesn’t know what they don’t know, so they can’t even ask the right question.

At VIA, we knew early on that private recovery phrases posed a significant usability and security issue for our users. In fact, our user research showed:

  • Users had no conceptual understanding of what a private recovery phrase was. In a SaaS model, your username and password gets you access to all your data, even if you switch devices. With cryptocurrency, you need your username, password, and private recovery phrase. Importantly, if someone has your private recovery phrase, they have access to your funds.
  • There are major usability hurdles for safely saving the private recovery phrase. Ideally, users would write the phrase down and store it in a safe place or save it in a password manager like 1Password. In practice, this is hugely impractical and makes major assumptions about the user’s situation and technical literacy.
  • Private recovery phrases posed a potential phishing risk to users. Some users thought of the phrase akin to a code you use to prove your identity to the bank. A scammer posing as a trusted organization could convince the user to give up the one piece of information that would give them full control over the users’ funds.

The damage? Lost funds.

The underlying cause? A system that didn’t support its users. In fact, it actively encouraged knowledge-based mistakes.

Remember Colin’s mantra: “no compromise.” Our solution needed to be both usable and secure.

VIA Wallet now has a feature called assisted recovery. This was particularly important for non-governmental organizations (NGOs) sending money to beneficiaries. Phones are frequently lost or stolen in high-need zones, leading to lost cryptographic keys and loss of digital assets. A BioCatch report from May 2025 highlights a massive surge in device theft across Latin America, noting that in Peru alone, over 1.7 million cell phones were reported stolen in a single year, averaging 4,676 thefts per day.

VIA Wallet users enroll as part of onboarding to VIA Wallet. With a single tap, the NGO is able to help the user recover their funds if they ever lose their device or get a new phone.

The NGO confirms the user’s identity, then VIA approves the NGO’s request. All requests are cryptographically approved. Each entity signs with their own credential. No one besides the user ever has access to their private recovery phrase.

The best part? We didn’t compromise on security.

This solution significantly reduces the chance of lost funds and does not force users to learn anything new. We stopped forcing users to adopt our mental model and instead matched the one they already have. Institutions that can verify the beneficiary’s identity and help them regain access.

Humans are humans, no matter how tech savvy or experienced. We all have moments when we are under pressure, are distracted, or sleep deprived. We all make mistakes and have slips and lapses. It is up to us as digital builders to anticipate these inevitable moments and to design safer systems for our users.