Quiz sprawdzający interakcje leków AI

Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.

1. How does a traditional rule-based interaction checker decide to fire an alert?
   • It predicts the next likely word in a sentence
   • It matches drug pairs against entries curated by clinical editors
   • It asks the patient to rate their symptoms
   • It searches social media for complaints
2. Why does the guide say a general chatbot can confidently describe an interaction that does not exist?
   • It is always connected to Micromedex
   • It generates statistically likely text rather than looking up a curated database
   • It deliberately exaggerates for safety
   • Interaction data is secret
3. Clarithromycin raising simvastatin levels is an example of which interaction type?
   • Pharmacodynamic
   • Pharmacokinetic
   • Food allergy
   • Placebo effect
4. An SSRI combined with tramadol raising serotonin syndrome risk is an example of what?
   • A pharmacokinetic interaction only
   • A food interaction
   • A pharmacodynamic interaction where effects add together
   • An interaction databases never include
5. What is the role of a research model like DeepDDI according to the guide?
   • It replaces pharmacy alert systems
   • It predicts possible interactions from drug features as hypotheses for study
   • It prints patient medication labels
   • It checks insurance coverage
6. Which known weakness of rule-based checkers leads clinicians to ignore alerts?
   • Są zbyt wolni, aby biegać
   • Alert fatigue from many alerts that get overridden
   • They cannot store drug names
   • Działają tylko offline
7. Why did the chatbot in the St. John's wort example miss the interaction?
   • St. John's wort has no interactions
   • The supplement was never included in the medication list it was given
   • Chatbots cannot read the letter S
   • The pharmacist hid the information
8. What does a retrieval-augmented design do differently from a plain chatbot?
   • It removes all citations
   • It first queries a curated database, then uses the language model to explain the retrieved entries
   • It trains a new model for every question
   • It relies only on the model's memory