Case study 06 · AI name origin

Where a name comes from,

and how sure the model is.

An AI explorer that predicts the most likely origin of a name: region, sub-region and country, each with a confidence score and a second-best answer. I designed it to show uncertainty honestly instead of hiding it behind one verdict. I led UX and UI end to end.

My role

Lead UX and UI designer

Scope

Research, IA, UX, UI, design-to-code

Platform

Web, English and Arabic

Sector

Public sector, under NDA

Name origin prediction result with region map, country and alternative country confidence gauges, sample data

01 · Snapshot

One name in,

a clear answer out.

Problem

A single answer hid the doubt

Names travel across countries and scripts. A model that names one country with no confidence invites too much trust.

What I did

Turned a prediction into an explanation

Search patterns, a broad to precise result layout, confidence gauges, ranked alternatives and a bilingual UI.

Outcome

Answers people can weigh

Every result shows where, how sure and what else, in one view.

3

levels of origin: region, sub-region and country

5

ranked alternative countries for every name

2

ways to search: full name, or first and last name

02 · The problem

One name, many spellings.

One answer is never enough.

The same name is written many ways across languages and scripts, and many names are common in several countries at once. A model that returns a single country with no confidence invites people to trust it more than they should.

First name · Last name · Full name · Spelling variants · Arabic and Latin script → Origin with confidence

01

Where is it from?

Region, sub-region and country, from broad to precise.

02

How sure is it?

A confidence score next to every answer, never hidden.

03

What else could it be?

A second country and a ranked list, not one verdict.

The question behind every search

Where is this name from, and how sure are you?

03 · The use case

From a typed name

to a clear answer.

A user types a full name, or a first and last name separately. The model compares it with known naming patterns and returns the most likely origin, from continent down to country, with how confident it is and what the alternatives are.

Detail view of a predicted name origin with ranked countries and regions, sample data

Screen · Detail view · Sample data

How a search flows

01

Choose how to search

Full name, or first and last name in separate fields.

02

AI predicts the origin

Region, sub-region and country, each with a probability.

03

Confidence in plain sight

A low to high gauge beside the main and the alternative country.

04

Explore the alternatives

Top countries and top regions, ranked by probability.

A plain prediction

One country, no reason

No sense of how sure the model is

Close second choices hidden

Searches lost after each lookup

With the explorer

Region, sub-region and country together

A confidence gauge on every answer

An alternative country and a ranked top five

Recent searches one click away

The model suggests an origin. People make the call.

04 · Key decisions

Four choices that keep

the AI honest.

01

Two ways to search

I saw

People hold names differently, and some only know part of one.

I chose

Full name in one field, or first and last name in two, chosen with a single switch.

Name origin search with full name or first and last name options
Origin summary cards for region, country and alternative country with probability gauges

02

Broad to precise

I saw

A weak country guess can still carry a strong regional signal.

I chose

Region, then sub-region, then country, each on its own map, so the answer is useful at every level.

03

Confidence beside every answer

I saw

A percentage on its own was easy to misread as certainty.

I chose

A low to high gauge sits next to the main and the alternative country, so 70% looks like 70%.

Detail view of a predicted name origin with ranked countries and regions, sample data
Ranked lists of top countries and top regions of origin by probability

04

Always an alternative

I saw

A single verdict hid close second choices.

I chose

A second country plus ranked top countries and regions make it clear the model suggests, it does not decide.

05 · Design details

Built to be read,

not decoded.

Each part of the result answers one question. Maps show where, gauges show how sure, ranked lists show what else, and the same view works in English and Arabic.

Sub region

Western Asia

Region · Asia

المنطقة الفرعية

غرب آسيا

المنطقة · آسيا

Components

Search switch · Name fields · Region map · Country map · Confidence gauge · Ranked list · Recent searches · Detail view

Accessibility

Colour plus words

Every map has a written label, so meaning never depends on colour.

Arabic

Mirrored, not translated

Layout, maps and gauges flip right to left with the language.

Clarity

One question per card

Where, how sure and what else, each in its own place.

06 · What it delivered

From a black-box guess

to an answer people can weigh.

3 levels of origin

region, sub-region and country in one view

Confidence on every answer

shown as a gauge, not a bare number

5 ranked alternatives

so close second choices are never hidden

2 languages

English and Arabic, with the layout mirrored

What I would do differently

Explain which part of the name drove the answer.

Showing whether the first or the last name carried the prediction would make each result easier to question, and easier to trust.

What I would keep

Show the doubt.

An answer with its confidence and a second choice is more honest than any single percentage.