Understand the technology
From evidence to a justified recommendation.
For May and November 2027. This lesson is for SL and HL students. Develop a recommendation step by step, using worked evidence, realistic alternatives and clear conditions.
01 | A RECOMMENDATION IS A DECISION WITH REASONS
Visionary Studios has investigated image generation and gathered test results. Someone now has to decide what the company should do. Listing every advantage and disadvantage does not, by itself, answer that question.
A recommendation proposes a course of action. A justification explains why that action is appropriate, using relevant evidence, company needs and acknowledged limitations.
A strong response might recommend a limited trial for concept development rather than immediate use for every final advertisement. Its strength comes from the reasoning, not from being the most enthusiastic or cautious answer.
02 | ANSWER THE ACTUAL DECISION
Identify what the question asks the company to choose: a model type, a workflow, a deployment approach or a next step in evaluation. These are different decisions.
If the question asks whether to expand a tested workflow, a long explanation of how diffusion works is not enough. Use technical understanding to explain the consequences of the proposed action.
Begin by completing this sentence: “For this purpose, I recommend that Visionary Studios…” Then check that the rest of the answer supports that decision.
03 | RETURN TO THE PRIORITIES
The studio needs useful creative outputs, but quality, consistency, resource use and responsible production can pull in different directions. Identify which matter most for the task being considered.
For a recurring character campaign, identity consistency may outweigh a small speed advantage. For early brainstorming, quick exploration may matter more than publication-ready precision.
Do not claim the booklet specifies your chosen priority unless it does. Explain your inference or state the assumption behind the recommendation.
04 | SEPARATE FACTS FROM PROPOSALS
Use three clear categories: facts supplied in the case, evidence from an explicitly described test, and proposals you are making.
“The studio achieved a 75% acceptance rate” is a factual claim requiring evidence. “A pilot could measure acceptance rate” is a proposed method. They are not interchangeable.
The numerical pilot below is fictional teaching material continued from 08.2. It must not be presented as a finding in the official booklet or an unseen examination scenario.
05 | MAKE AN EVIDENCE CHAIN
What was observed?
Why does it matter for the company?
What action does it support?
For example, repeated identity errors matter because designers must correct them before a campaign is approved. This can support limiting the workflow to exploratory concepts until consistency improves.
The middle step is essential. Quoting a number without explaining its consequence leaves the reader to supply the argument.
06 | REVISIT THE WORKED PILOT
| Measure | Workflow A | Workflow B |
|---|---|---|
| Candidates | 40 | 40 |
| Accepted candidates | 24 | 30 |
| Accepted proportion | 60% | 75% |
| Total production time | 240 minutes | 330 minutes |
| Minutes per accepted asset | 10 | 11 |
B has a 15-percentage-point higher accepted proportion. A uses one minute less per accepted asset. Neither result alone establishes the best choice for every task.
Before recommending either, inspect results by brief type and check whether essential requirements were met. The totals do not tell us which workflow preserved character identity better.
07 | EXPLAIN A TRADE-OFF WITHOUT INVENTING AN ADVANTAGE
If the priority is time per accepted asset, A has the better result in this pilot. If a higher proportion of suitable candidates is important, B has an advantage on that measure.
However, do not infer that B has better visual quality merely because more candidates passed. Acceptance under a minimum rubric is different from a comparison of quality among accepted images.
A careful recommendation names the advantage actually measured and avoids adding unsupported claims to make the chosen option sound stronger.
08 | CHECK ESSENTIAL CONDITIONS FIRST
An essential condition is a gate for proceeding, such as completing the agreed approval process. It should not automatically be overridden by a high performance score.
Suppose an important source concern remains unresolved for a proposed publication workflow. The studio should resolve or appropriately escalate it before proceeding under that plan, even if generation is fast.
Use a recommendation to explain how the gate will be addressed. Do not present a weighted total as if it settles every responsibility.
09 | COMPARE A PLAUSIBLE ALTERNATIVE
A justified choice explains why another reasonable option is less suitable for the stated purpose. The alternative might be a different generator, the existing human-led process or a combined workflow.
Avoid an artificial comparison where the rejected option has no benefits. For example, acknowledge that B produced a higher accepted proportion before explaining why A’s measured time advantage better matches a particular pilot objective.
This makes the conclusion a reasoned choice between options rather than a one-sided description.
10 | A CONDITIONAL RECOMMENDATION CAN BE DECISIVE
A conditional recommendation states an action and the conditions under which it should proceed. It is not the same as refusing to choose.
“Run a limited concept-art pilot with A, provided the agreed reviews are completed and client approval is obtained” is a definite next step. “It depends” without explaining what it depends on is not.
Choose conditions connected to identified uncertainties or requirements. Too many vague conditions can conceal the absence of a usable decision.
11 | MAKE IMPLEMENTATION REALISTIC
A recommendation becomes more practical when it identifies scope, responsibilities and a review point. Who will run the trial, assess outputs and decide whether to expand?
Start with a manageable production area, train the staff involved and preserve a suitable fallback workflow. Define what evidence would trigger continuation, revision or stopping.
These are proposed actions, not claims that the studio has already implemented them. Label them clearly when writing about the case.
12 | ACCOUNT FOR THE WHOLE COST
Per-image speed does not include every implementation cost. Setup, staff training, subscriptions, review, editing and maintenance can affect the benefit.
In a separate fictional planning example, suppose setup takes 120 staff-minutes and later work saves 2 staff-minutes per accepted asset. Ignoring other differences, 120 ÷ 2 = 60 assets are needed to recover that initial time investment.
This is a simplified time calculation, not a financial profit forecast. If savings vary or ongoing costs increase, the conclusion changes.
13 | ADDRESS THE PEOPLE AFFECTED
A studio decision affects designers, clients, creators and people represented in outputs. Explain relevant benefits and burdens rather than merely listing stakeholders.
For example, clients may gain more concept options while designers face additional review tasks. Training and realistic workload planning can help make the proposed change workable.
Connect source, fairness and disclosure measures to the earlier lessons. A good production result does not remove those responsibilities.
14 | USE UNCERTAINTY TO CHOOSE THE NEXT STEP
If a pilot has not tested peak demand or recurring characters, say so. That gap may justify further testing before expansion rather than a claim of universal suitability.
Distinguish “not demonstrated” from “cannot work”. Missing evidence is a reason to investigate, not permission to invent a result.
Prioritise the unknowns that could change the decision. A useful next test targets a significant uncertainty instead of repeating a favourable demonstration.
15 | CHECK HOW SENSITIVE THE CHOICE IS
A decision may depend heavily on one assumption, such as a small time saving or a chosen criterion weight. Ask whether a plausible change would reverse the preferred option.
If A’s advantage disappears once extra editing is included, the recommendation is fragile. If the same choice remains appropriate under several plausible conditions, confidence may be stronger.
This sensitivity analysis need not be complex. Explain the important assumption, vary it reasonably and describe the effect on the decision.
16 | COMPARE WEAK AND DEVELOPED REASONING
| Weak statement | Developed reasoning |
|---|---|
| “A is best because it is faster.” | “A used 10 minutes per accepted asset compared with 11 for B in the pilot. This supports a limited trial where production time is a priority, provided essential checks are met.” |
| “B is higher quality.” | “B had a higher accepted proportion, 75% against 60%. Further quality scoring of accepted images is needed before claiming superior visual quality.” |
| “Use AI for everything.” | “Limit the initial workflow to tested tasks and evaluate untested requirements before wider use.” |
Connectives such as because, however and therefore help express relationships when the reasoning behind them is clear.
17 | A WORKED RECOMMENDATION
For a fictional pilot objective prioritising time per accepted concept asset , I recommend a limited further trial of Workflow A. It used 10 minutes per accepted asset compared with 11 for B, which supports the studio’s aim of efficient concept production within this test.
However, B achieved a higher accepted proportion, and the aggregate results do not establish character consistency or quality differences among accepted outputs. Therefore the trial should inspect those requirements by brief type before expansion.
Proceed only after the agreed source, portrayal and disclosure reviews, with named approval responsibility. Retain the existing workflow for unsuitable tasks and review the decision after representative further testing.
This is one justified response under the stated priority, not an official answer or the only defensible choice. A different supported priority may lead to a different recommendation.
18 | WRITE YOUR OWN SUPPORTED DECISION
Recommend a next step for Visionary Studios using the fictional evidence. State the purpose, identify the strongest support, acknowledge an alternative and set a relevant condition.
Reveal a possible structure
“I recommend [specific action] for [purpose]. The evidence shows [accurate result], which matters because [company consequence]. Although [alternative strength or limitation], this choice is appropriate under [priority]. Proceed provided [essential condition], and collect [missing evidence] before [wider action].”
Use the structure to organise thought, not to replace subject knowledge. In an examination, use the supplied scenario and evidence. These independent notes and fictional pilot figures should be used alongside the official 2027 booklet.
CHECK YOUR UNDERSTANDING | CLASSIFY THE RECOMMENDATION ELEMENT