Evidence

Relevance

Evidence is relevant when it has any tendency to make a fact of consequence more or less probable than it would be without it (FRE 401; § 90.401). Relevant evidence is generally admissible and irrelevant evidence never is, but a judge may still exclude relevant evidence when its value is substantially outweighed by unfair prejudice, confusion, or wasted time (FRE 403; § 90.403). Relevance is the low bar every piece of evidence must clear before any other rule matters.

Hearsay

Hearsay is an out-of-court statement offered to prove that what it says is true (FRE 801; § 90.801). It is generally barred because the person who made the statement was not under oath and cannot be cross-examined, so its reliability cannot be tested the usual way. Both systems then carve out long lists of exceptions, such as the excited utterance and the business record, where the circumstances give a statement enough reliability to admit it anyway (FRE 803 and 804; § 90.803 and § 90.804).

Attorney-Client Privilege

This privilege protects confidential communications between a lawyer and client made to obtain or give legal advice, and it belongs to the client. The two systems locate it differently: in federal court, privilege is governed by common law that judges develop case by case (FRE 501), while Florida writes its privileges into statute, including the attorney-client privilege at § 90.502. Because it is a privilege, it can be waived, and disclosing the communication to an outsider usually waives it.

Work Product Doctrine

The work product doctrine shields materials a party or its lawyer prepares in anticipation of litigation, such as an investigator’s report or an attorney’s notes. It reaches more than communications, so it is broader than attorney-client privilege, but it is not absolute: an opponent can sometimes obtain fact work product by showing substantial need, while a lawyer’s mental impressions receive near-total protection. It comes from the discovery rules (Fed. R. Civ. P. 26(b)(3); Fla. R. Civ. P. 1.280(b)) rather than the evidence code, and students often confuse it with privilege.

Character Evidence and the Williams Rule

Character evidence is proof that a person has a certain trait, offered to argue they acted in line with it on the occasion in question, and it is generally barred for that propensity purpose (FRE 404; § 90.404). The same evidence of other acts may come in for a different purpose, such as proving intent or identity. Florida’s version of this other-acts rule is called the Williams rule after Williams v. State, and § 90.404(2) sets out how it may be used and the notice a party must give.

Expert Testimony and the Daubert Standard

An expert may give opinion testimony only if it rests on sufficient facts and a reliable method, and applies that method reliably to the case, with the judge acting as a gatekeeper against unreliable opinions (FRE 702). This is the Daubert standard, named for the Supreme Court decision that shaped it. Florida applied the older, more lenient Frye test for years, but its Supreme Court adopted Daubert in 2019, so § 90.702 now tracks the federal rule, a change that older sources often miss.

Authentication (Laying a Foundation)

Before an item such as a photo or a text message is admitted, the party offering it must show it is what they claim it is, a step called authentication or laying a foundation (FRE 901; § 90.901). This can be done through a witness with knowledge or the item’s own distinctive characteristics. Authentication is a separate hurdle from hearsay and relevance, so an item can clear one and still fail another.

Admissibility vs. Weight

Admissibility is a legal question the judge answers: do the rules permit the jury to see or hear this item at all? Weight is what the jury does next, deciding how believable or important the item is once it is admitted. When a lawyer argues that a problem goes “to weight, not admissibility,” they are conceding the evidence comes in and asking the jury, rather than the judge, to discount it.

Part 1: Read the Current Rules and Set Your Baseline

Start with the primary text, before you open any AI tool. On Cornell LII, read the federal rules on relevance (FRE 401 and 403), the hearsay definition and two common exceptions (FRE 801, plus the excited-utterance and business-records provisions in FRE 803), and expert testimony (FRE 702). Then open Chapter 90 on the Florida Legislature’s site and read the matching Florida sections: § 90.401 and § 90.403, § 90.801 and § 90.803, and § 90.702. Write down the actual language for a few of these. That wording is what you will hold the AI’s summaries up against later. One orientation point: Florida’s evidence rules sit in the statutes (Chapter 90), while Florida’s rules of civil procedure are a separate body of law hosted on The Florida Bar’s site, so do not expect to find the hearsay rule among the procedural rules.

Part 2: Spot the Issues Yourself

Read the fact pattern below. For each of the seven items, and before you use any AI, write a one- or two-sentence ruling in your own words. Is it relevant? Is it hearsay, and if so does an exception apply? Is it protected by privilege or work product? Does it raise a character or propensity problem? Does the expert opinion meet the reliability standard, and how would the item be authenticated?

Reyes v. Coastal Freight Inc. Dana Reyes sues Coastal Freight after one of its delivery trucks rear-ends her car at an intersection in Volusia County. She claims the driver was speeding and distracted. The following evidence is in play:

(a) A bystander shouted, “That truck ran the red light!” a few seconds after the crash.

(b) Coastal’s dispatch log, a routine GPS record, shows the truck traveling 55 mph moments before impact.

(c) A text the driver sent 30 seconds earlier reads, “almost there, running late.”

(d) An accident-reconstruction engineer Reyes hired will testify the truck was going 55 in a 35 zone, based on skid-mark analysis.

(e) A report by a private investigator, hired by Coastal’s lawyer, documents Reyes shopping and lifting bags after the crash, offered to suggest her injuries are overstated.

(f) An email between Reyes and her own attorney discusses her settlement strategy, and Coastal wants to use it.

(g) Records show the Coastal driver had two prior speeding tickets.

Once you have your own calls on paper, use the Socratic Foil prompt to have an AI question your reasoning one item at a time. The AI’s job here is to poke holes, so answer its questions and revise your analysis wherever it exposes a gap. Keep the thinking yours: if the AI simply asserts an answer, ask it why, and check that reason. When you are confident an item comes in, run the Weakest-Point Finder prompt on it to see the best argument against you before you commit to your ruling.

Part 3: Verify the AI and Find the Divergence

Now switch from reasoning to checking. Run the Rule Text Verifier prompt, pasting in the actual rule language you copied in Part 1, and ask the AI to apply only that text to two of the items. Compare its rule statements to the primary source word for word, and note any place it drifts from the text or states a rule from the wrong jurisdiction.

Then probe the one contrast at the center of this activity. Ask the AI how expert testimony is judged in Florida state court compared to federal court, and whether attorney-client privilege in each system comes from a statute or from judge-made common law. Watch the expert answer closely: many sources still say Florida follows the Frye standard, which stopped being true in 2019. Test whatever the AI tells you against the current text of § 90.702 and against In re Amendments to the Florida Evidence Code, 278 So. 3d 551 (Fla. 2019), which you can read on Google Scholar by selecting “Case law” and searching the citation. If the AI cites any other case, confirm on Google Scholar that the case exists and that it actually says what the AI claims. To organize the comparison, run the Jurisdiction Splitter prompt on one item and set the federal and Florida treatments side by side.

Part 4: Write Your Forum Post

Write 400 to 600 words reporting what you found. Cover which items in Reyes you would admit and which you would keep out, with the rule behind each call. Describe one point where you and the AI disagreed and who turned out to be right once you checked the text. Then give your results on the Florida expert-standard question and on any case the AI cited: was the rule stated correctly, and was the case real and on point? Close with a sentence on how far you would trust an AI’s unverified statement of an evidence rule in real work.

Part 5: Share Chat Link

Include one AI chat link with a 1–2 sentence explanation of what the conversation shows and why you chose to share it.

Socratic Foil

You are a law professor running a Socratic dialogue with me about evidence law. I am going to give you a civil fact pattern and my own analysis of whether each piece of evidence is admissible. Do not tell me the answers. Instead, take my items one at a time and ask me pointed questions that test my reasoning: whether a statement is hearsay and if an exception fits, whether something is relevant, whether it is privileged or work product, whether it raises a propensity problem, and how it would be authenticated. When my reasoning has a gap, ask a question that exposes it rather than filling it in. After I respond, tell me whether my revised answer holds up and why. Here is the fact pattern and my analysis: [paste].

This prompt combines role assignment with Socratic elicitation, and the instruction to withhold answers is the part that matters. Making the model question your reasoning instead of supplying conclusions keeps the analysis yours and surfaces the gaps you would otherwise skip past.

Rule Text Verifier

I am going to paste the exact text of specific evidence rules that I copied from Cornell LII and the Florida Statutes. Use only this text to analyze the evidence items I describe. Do not rely on your own memory of what these rules say, and if your general knowledge conflicts with the text I pasted, follow the pasted text and tell me where the conflict is. If applying a rule calls for a case, give the full citation and tell me plainly that I must confirm it on Google Scholar, since you may misremember case names or holdings. Here is the rule text, followed by the items: [paste].

This is source grounding: the model reasons from text you supply rather than from its training, which is exactly where fabricated rules and citations tend to appear. Asking it to flag conflicts and to mark every case as needing verification builds the checking step into the workflow instead of treating the model as an authority.

Jurisdiction Splitter

Compare how federal court and Florida state court treat one type of evidence: [expert testimony / attorney-client privilege / other-acts character evidence]. Put the answer in a table with a column for the federal rule and its citation, a column for the Florida rule and its citation, and a column noting whether the two currently agree or differ. State explicitly if the two systems used to differ but have since converged, and give the year of any change. Do not smooth over a difference that exists, and do not invent one that does not.

Structured output forces a complete, parallel comparison, because a table makes the model fill every cell for both jurisdictions rather than describing whichever one it finds easier. The prompt about convergence targets a common failure in which a model reports an outdated split because its training predates the change.

Weakest-Point Finder

Take one piece of evidence that I believe is admissible and argue the other side. Give me the single strongest objection a competent opposing lawyer would raise to keep it out, citing the specific rule. Then give me the best response I could make to that objection. Do not hedge or list every possible argument; identify the one with the best chance of working and explain why. The item and my position are: [paste].

Red-teaming asks the model to attack your position rather than affirm it, which tends to produce sharper analysis than a neutral review, since models drift toward agreement by default. Narrowing it to the single strongest objection forces a judgment about what actually matters, closer to how a lawyer prepares for an argument.