Documentation

General AI Knowledge

Documentation

Getting Started

•
Navigating EveryAnswer
•
Getting Started

Accessing EveryAnswer

•
System Requirements
•
Accessing Experts

Features and Functionalities

•
Workspaces
•
Experts
•
Multilingual Support

Expert Creation and Management

•
Expert Personas
•
Creating an Expert
•
Deleting an Expert
•
Expert Design and Branding
•
Conversation Starters

Data Management and Integration

•
Website Data Source
•
File Data Source
•
Questions and Answers Data Source
•
Forms
•
Supported File Types

Security and Privacy

•
Encryption
•
Privacy Controls

User Account Management

•
User Roles and Permissions
•
Creating and Managing User Accounts
•
Passwords

Sharing and Embedding Experts

•
Shareable Links
•
Embedding Experts on Your Website
•
Using Experts Privately

Billing and Subscription Management

•
Billing Cycles
•
Payment Methods
•
Invoices
•
Downgrading a Subscription
•
Cancelling a Subscription

Troubleshooting and Support

•
Getting Support

Introduction

In EveryAnswer, your Experts are designed to provide accurate and helpful answers based on the specific data you provide. However, there may be instances where your Expert doesn't have the necessary information to answer a user's question. In such cases, you have the option to dictate how the Expert should respond. This page will explain how to manage your Expert's responses when it doesn't know the answer, focusing on two key options and their implications.

Response Options for When the Expert Doesn't Know the Answer

When configuring your Expert, you can choose how it should respond when it encounters a question outside its training data. There are two primary options available:

Option 1: Answer Only from Provided Data (Default)

By default, EveryAnswer is set up to allow your Expert to answer only from the data you have provided. This ensures that all answers are based on verified and specific information from your documents, websites, and custom Q&As.

  • Provides answers rooted strictly in your data sources
  • Minimizes the risk of AI hallucinations or unpredictable answers
  • Ensures high reliability and relevance to your specific needs

Option 2: Use AI’s General Knowledge

You can also allow your Expert to leverage its vast general knowledge in cases where it cannot find an answer in the provided data. This can enhance the Expert's ability to handle a wider range of questions effectively, although it comes with some risks.

  • Expands the range of questions the Expert can answer
  • Introduces a higher risk of AI hallucinations or less accurate responses
  • Can be useful for addressing general inquiries beyond the specific dataset

Configuring Your Expert’s Response Settings

You can adjust these settings based on your needs. Follow these steps to configure how your Expert responds when it doesn't know the answer:

  1. Navigate to the Expert Settings in your EveryAnswer platform.
  2. Select the specific Expert you want to configure.
  3. Locate the section titled “How Should Your Expert Respond When it Doesn’t Know The Answer?”
  4. Choose between the two options:
    • Only allow the Expert to answer from the data you have provided (default)
    • Allow the Expert to answer from its vast general knowledge
  5. Click Save to apply your changes.

Implications of Choosing General Knowledge

While allowing the Expert to use its general knowledge can increase its ability to answer more questions, it's important to be aware of the potential trade-offs:

  • There's a possibility of the AI generating hallucinations or answers that are not as accurate.
  • Maintain awareness of this feature's experimental nature, and consider monitoring the Expert's responses closely if you enable it.

Maintaining High-Quality Responses

Irrespective of which option you choose, there are several strategies to ensure your Expert provides high-quality responses:

  • Regularly update and review the data you provide to the Expert.
  • Utilize custom Q&As to handle frequently asked questions or complex queries.
  • Take advantage of the escalation procedures to handle queries the AI can't satisfactorily answer.

By thoughtfully configuring your Expert and maintaining its data, you can maximize the effectiveness of EveryAnswer for your organization.

Related Documentation

Frequently Asked Questions

What are the two ways to configure how an Expert responds to unknown questions?
How Should Your Expert Respond When it Doesn't Know The Answer? You can configure your Expert to use only the data you provided or also refer to its general knowledge when the answer isn’t in the provided information. - Only allow the Expert to answer from the data you have provided: If the Expert encounters a question it cannot answer from the provided data, it will first check for any configured escalation procedures. If no escalation procedures are applicable, it will provide a response saying it can't answer the question. This ensures higher accuracy and reliability, as the Expert will only provide information that is verified and domain-specific. - Allow the Expert to answer from its vast general knowledge: If the Expert encounters a question it cannot answer from the provided data, it will first check for any configured escalation procedures. If no escalation procedures are applicable, it will try to provide an answer based on its general AI knowledge. This allows the Expert to respond to a wider range of questions, though it may increase the risk of hallucinations. Escalation Procedures: These are predefined rules that typically redirect the user to an appropriate person or resource based on the question. They may provide text, phone numbers, or links (URLs or email addresses) to ensure that user queries are appropriately addressed even if the initial Expert cannot provide an answer.
What does the default setting for responses to unknown questions allow?
The default setting only allows the Expert to provide answers based on the data you have supplied, including documents, public websites/knowledgebases, and custom Q&A's.
What factors should be considered when deciding to use the Expert’s general AI knowledge?
When deciding to use the Expert’s general AI knowledge, the following factors should be considered: - Accuracy: General AI knowledge may increase the risk of inaccuracies or hallucinations. Consider if the need for broad information outweighs the potential for errors. - Domain-Specific Needs: Evaluate if the information required is highly specialized, which may be better served by domain-specific data rather than general knowledge. - User Trust: Consider how using general AI knowledge might affect user trust in the Expert’s responses. Inaccurate or irrelevant answers could reduce confidence. - Escalation Procedures: Determine if there are robust escalation procedures in place to handle queries that the general AI knowledge cannot accurately address. - Context Sensitivity: Assess if the general AI knowledge can effectively understand and respond to the context of the questions asked. - Compliance and Security: Consider any compliance and security implications of using general AI knowledge, especially in sensitive or regulated industries.
Can you change the response settings after the initial configuration?
Yes, you can change the response settings after the initial configuration. This allows you to adjust how the Expert handles unknown questions, whether by restricting it to only use provided data or allowing it to refer to its general AI knowledge. You can also update escalation procedures to ensure the Expert remains accurate, reliable, and aligned with your organization’s needs.
How does restricting the Expert to only answer from provided data affect the reliability of the Expert?
Restricting the Expert to only answer from provided data significantly enhances its reliability. By ensuring responses are based solely on verified, domain-specific information, the likelihood of inaccuracies or hallucinations is greatly reduced. Additionally, when data is looked up from the provided data, we perform a quality check to ensure the answer is useful and relevant to the original question, and we conduct a fact check to ensure the response is supported by the information used to generate the answer. With this method, the risk of hallucinations is near zero, ensuring users receive accurate, trustworthy, and relevant information, leading to greater trust and confidence in the Expert’s responses.
What impact does allowing general AI knowledge responses have on the risk of hallucination?
Allowing general AI knowledge responses increases the risk of hallucination. Hallucination occurs when the AI generates information that is plausible-sounding but incorrect or nonsensical.
What does 'only allow the Expert to answer from the data you have provided' imply?
It implies that the Expert will only use the specific data you have supplied, such as documents and custom Q&As, to generate responses, rather than relying on its broader AI knowledge.

Related Documentation

No items found.
Last Updated:
October 8, 2024

Related Tutorials

Explore our step-by-step tutorials that provide detailed guidance on how to implement the concepts discussed in this document.
No items found.