Authentic Assessment Design Lab

Design the Task

Real-world authenticity 70%

Cognitive demand (Bloom's level) 65%

Rubric / criteria clarity 55%

Number of raters 24

Measurement Quality

Validity—

Inter-rater reliability—

Overall assessment quality—

Score std. dev.

0.0 pts

Raters within ±5

How it works

Authentic assessment (Wiggins, 1990) trades the easy consistency of a multiple-choice quiz for tasks that mirror real-world practice — but that trade has a measurable cost and benefit. This lab models both sides at once.

Validity — how well the task actually measures the transferable skill it claims to — rises as the task moves toward the "authentic zone": high real-world relevance and high cognitive demand (analysis, evaluation, creation, not recall):

d = √((A−0.9)² + (D−0.9)²) / √2
Validity = clamp(1 − 1.3·d, 0, 1)

Reliability — how consistently independent raters score the same performance — is helped by a clear rubric (C) but hurt by open-ended authenticity when criteria are vague:

Reliability = clamp(0.35 + 0.65·C − 0.25·A·(1−C), 0, 1)
score(t) follows an Ornstein–Uhlenbeck walk:
  dScore = θ(trueScore − score)dt + σ√dt·N(0,1),  σ = (1−Reliability)·25

Each glowing sphere is one rater's live score for the same fixed student performance (true score = 78). Tight clustering means high agreement; a wide, drifting cloud means the task is too open-ended for its rubric to pin down.

Overall quality is the harmonic mean of the two — a task only scores well if it is strong on both axes, not just one: Q = 2·V·R / (V+R). Real-world takeaway: an authentic task without a clear rubric looks rich but scores unreliably; a clear rubric on a rote task scores reliably but measures little that transfers.

Task position (authenticity × demand) Authentic-zone target Rater score cloud

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.