1.2 Computer-Adaptive Testing Mechanics & Technology-Enhanced Items

1.2 Computer-Adaptive Testing Mechanics & Technology-Enhanced Items

Quick Answer: FAST replaces fixed-form testing with a Computer-Adaptive Test (CAT) grounded in Item Response Theory (IRT). After each response the platform updates an interim ability estimate ($\theta$) and, guided first by the published blueprint and then by that estimate, selects the next item. Mathematics adapts item by item; ELA Reading adapts by passage set. A student must answer each item (or finish each passage set) before moving forward, and unanswered items count as incorrect - but FDOE's CAT FAQ is explicit that students may flag items with Mark for Review and may go back and change earlier answers. Alongside traditional multiple choice, FAST uses a defined catalog of Technology-Enhanced Items (TEIs), and the catalog differs between ELA Reading and Mathematics.

---

Item Response Theory (IRT) and the Computer-Adaptive Engine

To understand how the FAST assessment functions, candidates must understand the fundamental shift from Classical Test Theory (CTT) to Item Response Theory (IRT). Under CTT, tests are fixed-form: every student receives the exact same set of questions in the same order. A student's score is calculated simply by summing the raw number of correct answers. In contrast, FAST operates on modern psychometric IRT principles, specifically 2-Parameter Logistic (2PL) and 3-Parameter Logistic (3PL) models.

Under IRT, an individual test item is characterized not by its arbitrary placement on a test form, but by three invariant mathematical parameters:

1. Item Difficulty ($b$ parameter): The location on the ability scale where a student has a 50% probability of answering the item correctly. Items with higher $b$ values require greater ability to solve.
2. Item Discrimination ($a$ parameter): The slope of the Item Characteristic Curve (ICC) at the inflection point. A steep slope indicates that the question sharply differentiates between students whose ability lies just above or just below the item's difficulty threshold.
3. Pseudo-Guessing ($c$ parameter): The lower asymptote of the ICC, representing the statistical probability that a student with minimal ability will guess the correct answer on a multiple-choice item (typically around 0.25 for a four-option item).

                Probability of Correct Response P(theta)
  1.0 |                                 ...
      |                              .·
      |                            .·
  0.5 |--------------------------* (b = Difficulty inflection point)
      |                       .· |   (Slope at * = a Discrimination)
      |                    .·    |
  c   |..................·       |
  0.0 |__________________________|__________________
     -3           -1             0             +1           +3
                       Student Latent Ability (theta)

The Latent Ability Metric ($\theta$)

In IRT, a student's underlying proficiency is represented by the Greek letter theta ($\theta$), on a continuum roughly spanning $-3.0$ (very low) to $+3.0$ (very high), with $0.0$ near average. FDOE describes the starting point concretely: the first time a student takes a test in a subject the algorithm chooses a mid-level difficulty item; if the system already has prior data for that student in the same subject - from a previous grade or a previous PM window - the algorithm picks up where the student left off in terms of item difficulty. The stated exceptions, which always start mid-level, are grade 3 ELA Reading and Mathematics (students are moving up from the grade 2 assessments), grades 5 and 8 Science, the FAST ELA Reading Retake, and the EOC assessments.

Starting low is not a trap. FDOE addresses this directly: a test that begins with easier questions because of a weak prior window is still long enough to adapt upward, and a student with improved skills can move up one or more achievement levels within the same attempt.

[!IMPORTANT] The item bank is grade-level only. FDOE states that on grades 3–10 FAST tests, the algorithm does not reach above or below grade level - the bank for each test contains only items aligned to that grade's content. A struggling grade 7 student will not be served grade 5 items; they will be served the easiest grade 7 items.

Blueprint First, Then Fisher Information

As the student answers, the engine recomputes an interim $\theta$ using Bayesian or maximum-likelihood estimation. It then searches the secure, pre-calibrated bank of B.E.S.T.-aligned items. Two constraints govern the search, and the order matters: item selection is driven primarily by the test blueprint, which guarantees each student sees the required percentage of items from each reporting category, and only within those blueprint-eligible items does the algorithm pick the one carrying the most measurement information at the student's current $\theta$.

This is why blueprint coverage and adaptivity do not conflict. FDOE notes the practical consequence: blueprint coverage ensures each student sees a set percentage of items from each reporting category, but students will not necessarily see an item for every benchmark.

Fisher Information represents mathematical measurement precision. An item provides maximum information when its difficulty ($b$) matches the student's current ability ($\theta$). If a question is far too easy, a correct answer yields almost zero new information about the student's upper limits. If a question is far too hard, an incorrect answer yields almost zero information about the student's baseline competence. By continuously selecting items at the student's operational margin, FAST drastically reduces the Conditional Standard Error of Measurement (CSEM).

---

What Adaptive Delivery Actually Permits

This is the single most misunderstood part of FAST, and most third-party prep material gets it wrong. The rules below come from FDOE's Florida's Statewide Grades 3–10 and EOC Computer-Adaptive Tests (CAT) FAQ.

Two Different Adaptive Grain Sizes

Once the algorithm has selected the next passage set, that set is locked in - it does not change if the student goes back and edits answers inside the set they are currently working on.

                       [Student Answers Item]
                                 |
             -----------------------------------------
             |                                       |
       [Correct Answer]                       [Incorrect Answer]
             |                                       |
    [Theta Increases: +dTheta]             [Theta Decreases: -dTheta]
             |                                       |
  [Select Next Item with                  [Select Next Item with
   Higher Difficulty (b)]                  Lower Difficulty (b)]
             \                                       /
              ---> [Check Stopping Criteria] <-------

Skipping: Not Allowed, With One Nuance

On a linear paper test, students are coached to skip the hard one and come back. On FAST that does not work the same way, but the reason is narrower than most prep material claims.

• Mathematics: a student must provide an answer for each item to move on to the next item. The engine cannot select the next item without scoring the current one.
• ELA Reading: a student may move between items within a passage set without answering them, but must answer all items in the set before the platform will deliver the next passage set.

FDOE's own guidance is simply: students should always provide their best answer for each item as they encounter it. And under FAQ #17, items left with no response - including items never reached before the school day ends - are counted as incorrect. Leaving something blank is never a neutral act.

Going Back Is Allowed - and So Is Flagging

Here is where the common myth collapses. FDOE CAT FAQ #13 asks "Can students return to an item and change the answer?" and answers: "Yes. Students may return to items, and they are able to change the answer they provided if they choose to do so."

FAQ #14 explains why that is psychometrically safe: returning to an item and changing an answer does not affect the algorithm, because the items that follow have already been selected and delivered. The routing path is fixed once it happens. What a changed answer can affect is the score, because the final scale score is derived from the student's final response set.

FAQ #15 confirms the Mark for Review tool - flagging - is provided on all Florida computer-based tests. A student answers the item, selects Mark for Review from the item's drop-down menu, and a flag appears next to that item on the item review screen. Using it does not affect the algorithm or the score in any way.

[!IMPORTANT] The K–2 FAST assessments work differently: in the K–2 script, once a young student selects Next, they cannot go back. Do not carry that rule up into grades 3–10.

Pattern Scoring

FAST is scored by pattern scoring: the pattern of a student's answers is analyzed together with each item's statistics to determine scoring weights. The consequence is worth stating plainly - two students who answer the same number of items correctly may receive different scale scores. The student who handled the harder content in the bank scores higher. Guessing your way to a raw count is not a strategy.

---

Test Length and Psychometric Efficiency

CAT's headline advantage is length. The published FAST test lengths are:

Field-test items are embedded on the spring administration and do not count toward a student's score, which is why PM3 sessions run longer than PM1 and PM2. Students receive approximately the same number of items within a PM event, though a student's test can run a few items longer or shorter depending on how many field-test items they draw. The whole test fits in one session on one day, and FDOE recommends students take only one subject test per day.

Why can FAST achieve comparable or better precision with fewer questions? On a fixed form, many items carry almost no information about a specific student. A high-achieving student learns nothing about their ceiling from twenty easy questions, and a struggling student learns nothing about their floor from twenty questions far beyond reach. By concentrating items near each student's operating threshold, CAT raises precision across the whole ability range - and FDOE cites a second benefit: because students encounter different items, the tests are more secure.

---

Comprehensive Review of Technology-Enhanced Items (TEI)

Traditional multiple-choice questions encourage superficial elimination strategies and reward lucky guessing. To assess authentic critical thinking and rigor under the B.E.S.T. Standards, FAST utilizes an extensive array of Technology-Enhanced Items (TEIs) alongside standard multiple-choice items.

1. Multiple-Choice (Single-Select)

• Mechanics: Four traditional radio-button options (labeled A, B, C, D in review contexts, but presented without letters in the digital interface). Exactly one option is correct.
• Cognitive Target: Assesses direct factual recall, vocabulary in context, primary textual inference, or direct mathematical calculation.

2. Multiselect

• Mechanics: Presents five to six square checkbox options. The prompt explicitly instructs the student to select a specific number of correct choices (e.g., "Select the TWO sentences that best support the author's claim" or "Select ALL expressions equivalent to $3(2x + 4)$").
• Scoring Rules: Multiselect items typically enforce strict scoring. In items prompting for a specific number of choices (e.g., "Select TWO"), selecting three choices invalidates the answer and earns zero points. Partial credit is rarely awarded on standard operational items unless explicitly designated in blueprint rubrics.

3. Hot Text / Selectable Text

• Mechanics: Embedded directly within a literary or informational reading passage, or within a mathematical proof. As the student moves their cursor over the text, words, individual sentences, or entire paragraphs highlight in light blue. The student clicks on the highlighted block to select it as evidence.
• Cognitive Target: Eliminates artificial multiple-choice options by forcing the student to identify and extract textual evidence directly from the authentic passage.

4. Equation Editor

• Mechanics: Utilized throughout Grades 3–8 Mathematics. The interface presents a digital keypad containing mathematical operation symbols, fraction templates, exponents, radical signs, inequality symbols, and variables.
• Cognitive Target: Students construct mathematical expressions, equations, or numerical values from scratch. This eliminates distractor analysis and requires genuine computational fluency and procedural knowledge.

5. Table Match (ELA Reading) and Matching Item (Mathematics)

• Mechanics: Options are presented in columns and rows. The student clicks a box to match a correct option in a column with a correct option in a row. There is typically one correct option per row or column, though the number can vary. FDOE calls this Table Match on ELA Reading and Matching Item on Mathematics.
• Cognitive Target: Classifying shapes by attribute, sorting true/false mathematical statements, or matching claims across two informational texts.
• On paper-based accommodations: the item is modified so the student fills in a bubble.

6. Evidence-Based Selected Response (EBSR) — ELA Reading

• Mechanics: A two-part item. Part A is typically multiple choice and asks for an analysis or inference. Part B - multiple choice, multiselect, or selectable hot text - requires the student to point to the text that supports Part A.
• Cognitive Target: Claim plus evidence in a single scored unit. A confident answer in Part A that cannot be evidenced in Part B is a wrong answer.

7. Multimedia — ELA Reading

• Mechanics: Passages or items may include images, charts, graphics, or timelines. Any of the other item types can be used to assess the multimedia content.

8. GRID (Graphic Response Item Display) — Mathematics

• Mechanics: The student uses point, line, or arrow tools to build a response on a graph, or drags numbers, words, phrases, or images into a graphic.
• On paper-based accommodations: replaced with a different item type.

9. Fraction Model — Mathematics

• Mechanics: The student builds a fraction model. The tool either supplies a model already divided into equal parts or lets the student construct one by clicking + and − to set the number of wholes and the number of parts per whole, then shading parts.
• Cognitive Target: Elementary fraction reasoning without the crutch of answer options.

10. Graphing Item — Mathematics

• Mechanics: The student creates a graph or number line - a bar graph, line plot, or histogram; a plotted point; a graph of an equation; a shape on a coordinate grid; or an inequality graphed on a number line by choosing a point and an arrow.

11. Multi-Interaction Items

• Mechanics: Any of the above types can be combined into a single item with multiple parts, and each part can be a different type. On paper-based accommodations, selectable-response types may be combined into one item.

6. Editing Task Choice

• Mechanics: Presents a short passage with embedded inline dropdown menus replacing key words or punctuation marks. The student clicks the dropdown menu to select the grammatically correct word, verb tense, apostrophe placement, or transition word.
• Cognitive Target: Assesses language conventions, syntactic mechanics, and rhetorical coherence within authentic discourse rather than through disconnected grammar worksheets.

---

Strategic Navigation & Common Student Traps

[!WARNING]

Exam Trap 1: "I'll Just Leave It Blank and Come Back"

This is the trap that costs real points. In Mathematics the platform will not advance without a response, and in ELA Reading the passage set will not close without one. More importantly, FDOE scores unanswered items as incorrect. Answer every item with your best available choice, then use Mark for Review if you want to revisit it - flagging costs nothing and does not touch the algorithm.

[!WARNING]

Exam Trap 2: Multiselect Over-Selection and Under-Selection

When a multiselect question states "Select the TWO details...", selecting three choices immediately voids full credit on the item. Students must train themselves to re-read the prompt instruction and count their checked boxes before clicking "Next."

[!WARNING]

Exam Trap 3: Equation Editor Formatting Errors

In mathematics, entering an expression such as $\frac{3x + 1}{4}$ requires properly grouping the numerator inside the fraction template. Students who type 3x + 1 / 4 without using the fraction bar correctly create $3x + \frac{1}{4}$, which the computer scores as completely incorrect. Practicing with the official FDOE Equation Editor tool prior to test day is essential.

[!WARNING]

Exam Trap 4: Reading Isolation in Editing Tasks

In Editing Task items, students frequently read only the single sentence containing the dropdown menu. However, many items test paragraph-level verb tense consistency or global rhetorical transitions (e.g., However vs. Furthermore). Choosing an answer without reading the preceding and succeeding sentences leads directly into carefully designed distractor traps.