Duty of care — medical negligence
NSW 4 sources Common Law
JS
You
What is the standard of care required of a medical professional in NSW, and when does a failure to meet that standard constitute negligence?
E
Evatt

The standard of care for medical professionals in NSW is governed by both common law and the Civil Liability Act 2002 (NSW). A medical professional must exercise the degree of care, skill, and judgment that a reasonable person in that professional's position would exercise.

Key Principle
Under s 5O of the Civil Liability Act 2002 (NSW), a medical professional is not liable in negligence if they acted in a manner widely accepted by peer professional opinion as competent professional practice — even if contrary views exist. 1

The High Court in Rogers v Whitaker 2 modified the Bolam test for Australia: the standard of care is not solely determined by medical opinion — particularly for disclosure of risk. A doctor must warn of material risks a reasonable patient would attach significance to. 3

To establish negligence a plaintiff must prove on the balance of probabilities:

  1. The defendant owed a duty of care;
  2. There was a breach of that standard; and
  3. The breach caused the plaintiff's loss. 4
E
Evatt
Searching case law

Evatt Design System

Foundation tokens and components — Nordic minimalist · legal-grade precision

Color System

WCAG 2.1 AA on all text. No bright/aggressive fintech hues — black, off-white, deep navy, desaturated blue-green.

Ink
#0A0A0A
19.8:1 on Paper ✓
Paper
#F5F3EE
Background base
Navy
#0F1C35
18.2:1 on Paper ✓
Accent
#4A7C8C
4.6:1 on Paper ✓
Accent Pale
#E8EFF1
Surface tint
Muted
#8B8680
4.7:1 on Paper ✓
Border
#E0DDD6
Dividers only
Error
#8B3A3A
4.8:1 on Paper ✓

Type Scale

Playfair Display — headings, hero, editorial. Inter — all UI, body, labels. JetBrains Mono — citations, tokens.

display/4xl64pxLegal Intelligence
display/3xl48pxLegal Intelligence
heading/xl28pxSection Heading
heading/lg21pxSub-section Heading
body/md17pxBody text for readable paragraphs and legal analysis copy.
body/base15pxUI body text, chat messages, standard interface copy.
label/sm13pxLabels, metadata, secondary information
caption/xs11pxUPPERCASE LABELS · BADGES · TAGS
mono13px[2024] HCA 14 · s 5O CLA · --sp-4

Spacing Tokens

4px base unit. All layout decisions on 8-point grid.

--sp-14px
--sp-28px
--sp-312px
--sp-416px
--sp-520px
--sp-624px
--sp-832px
--sp-1040px
--sp-1248px
--sp-1664px
--sp-2080px

Components

Button System — 4 variants, 3 sizes

Buttoncomponent/Button

Citation Authority Badges

Badge — citation statuscomponent/Badge
Binding Persuasive Overruled ✓ Verified

Citation Card — expandable source reference

CitationCardcomponent/CitationCard
1
Rogers v Whitaker
(1992) 175 CLR 479
"The standard of care is not determined solely by the practice of the medical profession…"
Binding High Court 1992
✓ Verified in database

Chat Components

Message bubbles, states, and interaction patterns used in the chat interface.

User Message

MessageBubble / usercomponent/Message
JD
You
What is the duty of care standard in medical negligence under NSW law?

AI Response with Inline Citations

MessageBubble / assistantcomponent/Message
E
Evatt AI

Under s 5O of the Civil Liability Act 2002 (NSW), a medical professional is not liable in negligence if they acted in accordance with peer professional opinion widely accepted as competent practice. 1

However, the High Court in Rogers v Whitaker 2 qualified this for disclosure of risk — the standard is not solely determined by medical opinion where a patient would attach significance to a risk.

Key Principle Block

KeyPrinciplecomponent/KeyPrinciple
Key Principle
A medical professional's duty of care is not discharged by conforming to peer opinion alone — patients must be warned of material risks they would reasonably attach significance to. 2

Streaming State

StreamingIndicatorcomponent/StreamingIndicator
E
Searching case law
E
Verifying citations
E
Composing answer

Inline Citation Chips

CitationChipcomponent/CitationChip
Inline chips appear within prose at the point of claim. 1 Multiple sources can be stacked 23 without breaking reading flow.
States: 1 2 3

Message Actions

MessageActionscomponent/MessageActions
Appears below assistant messages on hover. Minimal — no icons, no colour, no noise.
AI Legal Research Platform

Research faster.
Cite with
confidence.

Evatt surfaces binding case law, statutes, and secondary sources — every claim verified, every citation traceable. Built for lawyers who don't tolerate hallucinations.

What is the standard of care for medical professionals?
Standard of care for medical professionals in NSW?
Under s 5O of the Civil Liability Act 2002 (NSW) 1, a professional does not incur liability if they acted in a manner widely accepted by peer professional opinion. The High Court in Rogers v Whitaker 2 modified the Bolam test for Australia…
Sources
1
Civil Liability Act 2002 s 5O
2
Rogers v Whitaker (1992) 175 CLR 479
3
Adeels Palace v Moubarak (2009) 239 CLR 420
0%
Hallucination rate on verified citations
Faster than manual legal research
¼
The price of Harvey or Thomson Reuters
2.4k+
Lawyers at top-tier firms

Issues Identified & Solutions

Issue 01
Headline "Making Lawyers Lives Easier" is generic. Reads like every AI SaaS product.
Replaced with "Research faster. Cite with confidence." Specific, outcome-focused, speaks directly to the lawyer's core anxiety (hallucinated citations). Subhead addresses the hallucination fear explicitly.
Issue 02
No social proof or trust signals above the fold. Critical for a sceptical legal audience.
Added lawyer count + firm tier signal beneath CTA. Added stats strip (0% hallucination rate, 4× speed, ¼ price). Concrete numbers lawyers can evaluate quickly.
Issue 03
Hero visual doesn't show the product. Abstract backgrounds don't build trust with detail-oriented lawyers.
Replaced with live product mock showing an actual AI response with inline citation chips and verified sources panel. The key differentiator (citation transparency) is visible before any scroll.
Evatt AI — Take Home Assignment

Design Decisions

A written account of every significant design decision across the chat interface, design system, and website redesign, with the reasoning behind each choice and how I'd validate them.

01

Chat Interface Redesign

Decision A: Three-column layout

The interface is split into sidebar (conversation history) / main chat / citation panel. This mirrors how a lawyer actually works: they keep their research trail visible while reading the AI's answer, and they want sources immediately accessible without scrolling or clicking away.

Validation metric: Time-to-first-citation-click. If lawyers find the right source faster than in a two-column layout (sidebar + chat only), the three-column model is validated. Target: <8s to locate a cited case.

Decision B: Inline citation chips [1] [2]

Citations are shown as small numbered chips inline within the AI's prose, not as a footnote list appended at the bottom. This keeps reading flow intact: the lawyer sees which claim is backed by which source at the exact moment they read it, rather than needing to cross-reference later.

Validation metric: Misattribution rate in user testing. Ask lawyers to identify which case supports a given claim. Inline chips should outperform footnotes on accuracy and speed.

Decision C: "Key Principle" highlight block

When the AI identifies a leading case or core statutory rule, it is surfaced in a tinted block with a left accent stroke rather than buried in the third paragraph. Lawyers are trained to identify the ratio decidendi. This mirrors that mental model. The block says "this is the rule you'll cite; everything else supports it."

Validation metric: Comprehension accuracy: does the lawyer correctly identify the controlling authority after reading the response? Compare highlight block vs. plain paragraph.

Decision D: Streaming indicator labels source activity

Rather than a generic spinner, the streaming indicator says "Searching case law", giving the lawyer confidence that something substantive is happening, not that the system is frozen. For a sceptical audience, perceived AI process transparency reduces drop-off during latency.

Validation metric: Session abandonment rate during response generation. Labelled indicator should show lower abandonment vs. unlabelled spinner.

02

Design System Foundation

Decision: Palette: navy + off-white + desaturated teal

The palette deliberately avoids bright saturated colours common in fintech (orange/red on black). Law is a conservative profession; the palette signals seriousness, institutional trust, and restraint. Deep navy (#0F1C35) reads as authoritative without being oppressive. The off-white paper tone (#F5F3EE) is warmer than pure white, reducing eye strain during long research sessions.

Decision: Inter for UI, Playfair Display for headings

Inter is the standard for high-legibility professional UI. Playfair Display brings an editorial, almost journal-like quality to headings, signalling this is a research tool, not a consumer chatbot. The combination creates hierarchy without needing weight contrasts alone: serif for "what this section is," sans for "what to do."

Decision: Citation authority badges as first-class component

Binding / Persuasive / Statute / Overruled are not decorative labels; they are legally meaningful distinctions a lawyer must assess before relying on any source. Making them a distinct, colour-coded component (not just text) means lawyers can scan a citations panel and triage authority in under two seconds. The Overruled badge uses the error colour to signal "do not cite this" at a glance.

03

Website Hero Redesign

Issue 01 identified
Headline "Making Lawyers Lives Easier" is generic

Every AI SaaS product says some version of this. It says nothing about how Evatt is different from Harvey, CoCounsel, or a Google search.

Fix: "Research faster. Cite with confidence." Specific, outcome-focused, directly addresses the lawyer's core anxiety: hallucinated citations that could embarrass them in court or with a client.
Issue 02 identified
No trust signals above the fold

The existing hero has no social proof, no firm-count, no credibility anchors. Lawyers are among the most sceptical adopters of new technology; a blank hero with a CTA asks them to trust a brand they've never heard of.

Fix: Added lawyer count ("2,400+ lawyers at top-tier firms"), a stats strip (0% hallucination rate, 4× faster, ¼ the price), and an avatar cluster beneath the CTA. These are concrete claims lawyers can evaluate, not vague superlatives.
Issue 03 identified
Hero visual doesn't show the product

The original hero uses an abstract/decorative visual. For a detail-oriented legal audience, abstract imagery signals "marketing" not "tool." They want to see what they're buying.

Fix: Right panel shows a live product mock: an actual AI response with inline citation chips and a verified sources panel. The key differentiator (citation transparency) is visible before any scroll. A lawyer can evaluate the product in 3 seconds without signing up.
📐
Success metric for hero redesign

Primary: free trial sign-up conversion rate. Secondary: scroll depth (did the user reach the stats strip, indicating engagement beyond the headline). Both measurable via a simple A/B test between the old and redesigned hero.

04

Instagram Carousel

Decision: Slide narrative arc: provocation → proof → price → CTA

Slide 1 opens with a bold competitive provocation ("Forget Harvey. Forget the rest.") to stop the scroll. Slide 2 delivers product substance (feature list). Slide 3 makes the price comparison concrete. Slide 4 closes with a CTA. This mirrors a sales conversation: hook → credibility → objection-handling → ask.

Decision: Alternate dark/light slides

Navy → paper → navy → navy creates visual rhythm and signals "there is more here" on the first swipe. The light slide 2 (features list on off-white) deliberately uses the product's UI aesthetic; the carousel IS the design system applied to social, not a separate visual identity.

Decision: No stock photography, no illustrations on carousel

Legal imagery (Lady Justice, gavels, leather books) is clichéd and sends "old firm" rather than "modern tool." The carousel uses typography, grid, and the product's colour tokens as its visual content. Clean, confident, consistent with the brand's Nordic minimalism, and cheaper to produce at scale.

Overarching design philosophy

"A serious research environment with modern ergonomics, not a consumer chat app."

Every decision in this file flows from one constraint: the user is a time-poor, detail-oriented lawyer who is professionally liable for the advice they give. The UI must earn their trust on first sight, make verification effortless, and stay out of the way of their judgment. Evatt amplifies the lawyer; it doesn't replace them.

Loom Walkthrough — 3–5 min

Speaker Notes

One card per section. Timings are guides. Speak to the screen — don't read verbatim.

Section 00 ~30 sec
Intro
  • "I'm walking through my take-home for Evatt — four tasks: chat interface redesign, design system, website hero, and an Instagram carousel."
  • "Everything is built as a live prototype — I'll show you the actual screens, then explain the decisions behind each one."
  • Switch to tab 01 — Chat Interface.
Section 01 — Chat Interface ~75 sec
Chat Interface Redesign
  • "Three-column layout: sidebar for history, main chat, citations panel. Mirrors how a lawyer actually works — they never want to leave the page to verify a source."
  • "Inline citation chips [1] [2] appear at the exact claim, not as footnotes. The lawyer doesn't have to scroll down to cross-reference."
  • "The Key Principle block surfaces the ratio decidendi — the rule the lawyer will actually cite — so it's not buried in paragraph three."
  • "Streaming label says 'Searching case law' rather than a blank spinner — transparent AI process builds trust with a sceptical audience."
  • Switch to tab 02 — Design System.
Section 02 — Design System ~45 sec
Design System Foundation
  • "Navy, off-white, desaturated teal — no bright fintech colours. Law is a conservative profession; the palette signals institutional trust."
  • "Playfair Display for headings gives an editorial, journal-like quality. Inter for UI. The combination signals 'research tool' not 'consumer chatbot.'"
  • "Authority badges — Binding, Persuasive, Overruled — are a first-class component, not just text. A lawyer can triage an entire citations panel in under two seconds."
  • Switch to tab 03 — Hero Redesign.
Section 03 — Hero Redesign ~60 sec
Website Hero Redesign
  • "Original headline: 'Making Lawyers Lives Easier' — generic, says nothing. Replaced with 'Research faster. Cite with confidence.' Speaks directly to the lawyer's core anxiety: hallucinated citations."
  • "Added a stats strip: 0% hallucination rate, 4x faster, quarter the price. Lawyers are sceptical — they need concrete numbers, not superlatives."
  • "Right panel shows the actual product: a live AI response with inline citation chips. The key differentiator — citation transparency — is visible before any scroll."
  • Switch to tab 04 — Instagram Carousel.
Section 04 — Carousel ~45 sec
Instagram Carousel
  • "Narrative arc: provocation, proof, price, CTA. Slide 1 stops the scroll — 'Forget Harvey. Forget the rest.' Slide 4 converts."
  • "Slide 3 is a positioning chart: intelligence vs affordability. Evatt sits top-right — high capability, low price. Competitors bottom-left."
  • "No stock photography — the design system tokens are the visual content. Cheaper to produce, consistent with the brand."
~30 sec
Close
  • "Every decision here flows from one constraint: lawyers are time-poor, detail-oriented, and professionally liable for what they cite."
  • "The design has to earn their trust on first sight, make verification effortless, and stay out of the way of their judgment."
  • "Happy to go deeper on any of these — thanks for watching."
Timing summary
~30s
Intro
~75s
Chat
~45s
Design System
~60s
Hero
~45s
Carousel
~4m
Total