Improper

Podcast 2 — In conversation

Speaker A: I had a situation at work where someone sent a message to the whole team that felt off — slightly personal, slightly unprofessional. My manager described it as improper communication. Is that the right word?

Speaker B: It is precise, yes. Improper communication in a professional context suggests a violation of expected conduct — not necessarily anything dramatic, but something that crossed a boundary of what is considered acceptable in that setting. It is a fairly formal choice, which makes it suitable for HR language or a written warning, but a bit stiff for casual conversation with a colleague.

Speaker A: I see — so the formality of the word itself signals something?

Speaker B: Exactly. When someone uses improper rather than just wrong or inappropriate, they are usually operating within a formal framework — a workplace policy, a legal standard, a code of conduct. The word implies there is a defined norm that has been breached. That is slightly different from inappropriate, which is more situational and less tied to a codified rule. Improper often implies that someone should have known better — that the standard was clear.

Speaker A: What about the mathematical sense — an improper fraction? That is a completely different context.

Speaker B: Yes, and it is a good illustration of how the word works technically. In mathematics, a proper fraction has a numerator smaller than the denominator — like three over five. An improper fraction has a numerator equal to or greater than the denominator — like seven over four. It is not wrong or bad in any way; it simply falls outside the definition of a proper fraction. That technical use strips away the moral dimension entirely and shows the word in its purest structural sense: not fitting the defined category.

Speaker A: Is there a difference between improper and indecent? I sometimes see those used together.

Speaker B: They overlap in the social register, but indecent is considerably stronger. Indecent carries explicit connotations of offensiveness, often related to sexual conduct or public behaviour — indecent exposure, indecent language. Improper is more restrained; it describes a departure from propriety or protocol without necessarily implying offence at the same level. An improper remark might be tactless or presumptuous. An indecent one is more directly offensive. In legal and formal use, improper conduct is the more common and measured term.

Podcast 3 — Prompt engineering

Speaker B: Today we are working with improper in development prompts. This is a precise and powerful word for prompting because it tells the AI to identify, avoid, or correct behaviour that falls outside defined standards — whether those standards are technical specifications, coding conventions, UI guidelines, or business rules. Improper is more forceful than incorrect and more specific than wrong. Let me show you how it performs across different scenarios.

Speaker A: I like that distinction. Improper implies a standard exists and has been breached. Where do we start?

Speaker B: Let us open with a UI design prompt where improper flags a usability standard violation.

Speaker B: PROMPT: Audit this navigation bar design and identify any improper accessibility implementations. Flag all instances where interactive elements lack visible focus indicators, where colour contrast ratios fall below WCAG AA standards, where click targets are smaller than 44 by 44 pixels, or where screen reader labels are absent or improper. Provide corrected markup for each improper element found.

Speaker B: Using improper twice there — improper implementations and improper labels — tells the AI to evaluate against a recognised standard, not just general preference. The output becomes a compliance audit, not a style opinion. That level of specificity is very hard to achieve with vague words like bad or wrong.

Speaker A: That is a sharp prompt. Using improper twice reinforces the standard-based framing. What about database work?

Speaker B: Here is a schema review prompt using improper.

Speaker B: PROMPT: Review this PostgreSQL schema for an e-commerce platform and identify any improper normalisation decisions. Flag tables where repeated data groups suggest improper first normal form violations, where non-key attributes depend on only part of a composite key indicating improper second normal form, and where transitive dependencies exist indicating improper third normal form. Suggest corrected table structures for each improper design.

Speaker B: Framing normalisation violations as improper decisions gives the AI a precise technical lens. It is not asking for general feedback — it is applying a formal set of rules and asking for deviations from those rules to be surfaced.

Speaker A: Tying improper to the formal normal form rules is very direct. What about application logic?

Speaker B: Here is a business logic validation prompt.

Speaker B: PROMPT: Analyse this leave approval workflow in an HR management system and identify any improper state transitions. Flag cases where a leave request can move from Approved directly to Draft without passing through Cancelled, where improper role assignments allow employees to approve their own requests, and where absence of audit logging creates improper accountability gaps. Produce a corrected state machine diagram with each improper transition removed.

Speaker A: Improper state transitions and improper role assignments — that is extremely tight framing for a workflow audit. What about a full application build prompt?

Speaker B: Here is one.

Speaker B: PROMPT: Build a staff expense submission module for an accounting system. Reject any submission where the amount field contains improper input such as negative values, non-numeric characters, or values exceeding the per-category policy limit. Flag improper receipt attachments where file type is not PDF or JPEG or file size exceeds five megabytes. Display a clear inline validation message for each improper field before allowing submission.