Miraai · Reasoning

The item exists only
while you hold the orb.

Three example items, in the test's own interface: one verbal, one numerical and one abstract. Nothing here is scored and nothing is recorded; these are the practice items every candidate sees, and the live bank is not on this page.

The orb falls and settles. The screen is otherwise blank. Press anywhere and the orb is drawn to your pointer — you never have to catch it. The clock starts when it arrives.
Let go and the item disappears, but the clock keeps running. Press anywhere to call the orb back and the item returns.
Answer by carrying the orb onto one of the response areas. Released anywhere else, it waits there until you place it.
Watch the panel beside the test. It is the interaction trace the test records: the path your pointer takes, the movement features read from it, and every release, tab switch and paste.

01

Welcome. Here is your first section.

Section one of three

Verbal Reasoning

—

One example item, which shows you the answer and the reasoning behind it.

That is the practice done. Good luck.

—

Work on your own. No other person, no second device, no website, and no AI assistant. The orb is there to keep this honest: the item is hidden the moment you let go.
No paper, pencil or notes. Everything you need is on the screen.
One item at a time. You cannot go back to an item once you have answered it.
Need a break? Don't pick the orb up on the next item, or simply close the browser. Your place is saved and you will carry on from the next item.

How this assessment works

Instructions

Hold the orb to see the item. Press anywhere and the orb comes to you. Release it and the item disappears while the clock keeps running, until you call the orb back.
Answer by carrying. Bring the orb onto your chosen response and let go there. Released anywhere else, it waits until you place it.
Each item has its own time. Harder items are given more of it, from about 20 seconds at the easiest level to 75 at the hardest. The allowance for the current item is shown at the top right.
Work on your own. No other person, no second device, no website, no AI assistant, and no paper, pencil or notes.
You are not told whether you were right. Only the practice items show the answer, as they do here.

That is the mechanic

The question is only there while you hold it.

In a live sitting the items keep coming, each one chosen from the response before it, and the trace you just produced is read alongside the answers. Nothing from this page was scored or stored.

Assessment complete

—

Interaction Authenticity Index

—

The composite is a unit-weighted mean of the three scale θs, restandardised to the reporting metric assuming an inter-scale correlation of 0.55. Measurement errors are independent across separately-administered scales, so the composite is more reliable than any scale in it. Differential weighting needs a validated factor structure or criterion data, and until you have either, unit weights are the defensible choice. Reporting is floored at θ = −2.50: below that the scale carries no discrimination, so everyone reports STEN 1 and we say so rather than implying a difference that is not there.

The IAI in this prototype is an illustrative heuristic, not a validated classifier. In production it is a logistic model trained on labelled human and synthetic sessions, and it returns a review flag with reason codes, never an automatic determination. No candidate should be failed on telemetry alone.