Tuning FAQ
This page combines frequently asked questions about tuning AI responses and managing feedback and FAQs.
Tuning Responses
What is the importance of keeping the knowledge base updated for Alhena AI?
Updating the knowledge base ensures that the AI remains effective and relevant, avoiding the "garbage in, garbage out" problem by only using clear, accurate, and current information.
How can human feedback be used to improve Alhena AI's responses?
Human feedback allows admins to correct or enhance AI responses directly. This targeted feedback is crucial when the AI provides incorrect or unsatisfactory answers, helping it to learn and adapt for future interactions. Refer to this guide on adding human feedback.
What are AI Personality and Guidelines in Alhena AI?
This feature lets users customize the AI's tone, name, and guidelines for responding. Guidelines are rules that influence how the AI Agent behaves and responds in specific scenarios. For more information, refer to this guide.
When should I use human feedback instead of setting guidelines?
Use human feedback for immediate corrections or improvements on specific answers. Guidelines should be set for standardizing responses across broader scenarios or implementing consistent behavioral rules.
Can I customize the identity and tone of Alhena AI?
Yes, the AI Agent's identity and tone can be customized in the "Personality" section of the dashboard, allowing the AI Agent to align more closely with your company's branding or communication style.
What should be considered before adding a new guideline?
Decide if you really need to add a guideline or providing human feedback will suffice. Consider whether the guideline overlaps with existing functionalities or is beyond the AI's capabilities, ensuring it's necessary and will effectively enhance the AI Agent's performance.
How can I ensure the guidelines I set are effective?
Ensure guidelines are actionable, concise, and directly influence the AI Agent's interactions. Avoid overloading the AI Agent with too many rules, which can confuse and diminish its effectiveness.
Should I update the AI knowledge base or should I add Human Feedback?
It is important to note that it might easier to add Human Feedback and the AI learns from the feedback in real time. But having an outdated or incorrect knowledge base will always play a role in confusing the AI. Although the AI gives high priority to Human Feedback in case of conflicting knowledge base.
If the answer of AI is incorrect while the knowledge base is correct, adding Human Feedback will fix the problem.
I am getting "Sorry I couldn't find an answer to your question" as an answer to my questions. What should I do?
If you encounter this response, take the following steps:
Check Answer Sources: Verify if the dashboard's Answer Sources contain the necessary information. If not:
Update the Knowledge Base: Add or update documents with the correct, clear, and relevant information.
Provide Human Feedback: If the knowledge base is accurate but the AI fails to answer, use the "Provide Answer Feedback" option to guide the AI on responding to similar future queries.
Review AI Capabilities: Ensure the query is within the AI's capabilities and not restricted by guidelines. Adjust guidelines as needed.
Security Measures: If attempts to trick or hack the AI are detected, the user will be blocked. If this was in error, please contact our support team.
The AI gave an incorrect answer, how do I fix that?
To correct an incorrect answer given by the AI, use the human feedback tool. Navigate to the dashboard, find the question under "Conversations," and use the "Provide Answer Feedback" option to submit the correct response or guidance. This direct intervention helps the AI learn and improve its future responses. Follow this guide.
How can I fix the case when AI provides a link in the answer which was 404?
It is possible that the link is removed but is being referenced in other documents. You should update the knowledge base to remove or replace the outdated link.
Additionally, providing human feedback for that specific answer can immediately correct the response, guiding the AI not to use that link in the future.
I want the AI to avoid answering certain type of questions
To prevent the AI from answering certain types of questions, you can set specific guidelines in the AI Personality and Guidelines section of the dashboard. This is an example guideline: If your answer has potential of steering financial decisions, mention the disclaimer - "This is not financial advice. Please do your own research before making any financial decisions.". Another Example: Never provide investment advice or discuss strategies for making money or getting rich.
I have added some guidelines but the AI doesn't follow them
If the AI is not following the guidelines you've set, it could be due to unclear or conflicting instructions. Ensure that the guidelines are clear, concise, and do not conflict with the AI's existing guidelines. Follow this guide to write effective guidelines.
Can I see what part of a document was used by Alhena AI to generate the answer
Identifying the specific part of a document that Alhena AI used to generate an answer can be challenging. This is because the AI synthesizes information from multiple document chunks (up to 20 relevant pieces) to construct a response. As a result, any single paragraph in the AI's response may be a hybrid, incorporating elements from several different sources. Although you can view the 'Answer Sources' in the dashboard to get a general sense of which documents influenced the AI's response.
Can I edit the Human Feedback that I provided for an answer?
Yes, You can easily edit your provided Human Feedback.
Search the Conversations to find the question for which you had provided Human Feedback. Alternatively, you can manage previous feedback from the FAQs page.
From the conversations screen, you'll see messages that have previous feedback clearly marked,
You can then see the FAQ that was added to your knowledge base based on this feedback. From here, you can manually make any changes to the FAQ that was generated based on your feedback, or delete the FAQ. If the FAQ is deleted, you'll be able to submit new feedback again from the conversations page.
Adding Feedback and FAQs
Q: Is there a way to provide AI feedback without it turning into a new FAQ?
By default, any feedback you provide on an AI generated answer creates a new FAQ. This FAQ helps the AI respond better to similar questions in the future.
If your feedback is a general instruction rather than a specific correction, you should add it directly to the AI's Answering Guidelines. Guidelines serve as broad instructions instead of specific FAQs. For example, if you want the AI to always greet users at the start of a conversation, add this as a guideline rather than as an FAQ.
Q: How can we keep FAQs updated when our documentation changes frequently (weekly or more)?
Currently, keeping FAQs updated is a manual process. When reviewing conversations, you'll see the exact FAQ used by the AI as the source for each answer. If you notice an outdated answer, you can directly click on the relevant FAQ and edit it to reflect the latest information.
Q: Can FAQs include media (GIFs, images, etc.) so the AI can send rich replies?
Yes, in the answer feedback, you can suggest adding relevant images or GIFs. The AI will incorporate these into the FAQ it generates. You can review the generated FAQ, and if the AI hasn't included the media you intended, you can manually add the media URL by editing the answer.
Q: In the dashboard, are FAQs sorted by created date or last-modified date?
FAQs are sorted by last-modified date.
Q: Can we assign an FAQ ID/number so each entry is easy to spot without reading the whole text?
Currently, this feature is not available.
Q: Is there a way to bulk-export all FAQs (or just their links) to CSV/Excel for offline review?
Not directly through the Alhena dashboard, but you can contact the Alhena support team, and we'll provide you with a CSV containing all FAQs.
Q: How does the FAQ system handle "hints" — answers that point to an article rather than contain the full instructions?
When providing feedback on an answer, you can include the title of the article you'd like the AI to reference. The AI will then review the referenced article when generating the FAQ.
Q: What exactly is the Answer field in an FAQ? Is it conversation-specific, or does it become a global reply once saved?
Once saved, the Answer field becomes a global reply. During the initial FAQ creation, the AI considers the specific conversation context, but after saving, the answer is conversation-agnostic and available globally.
Q: I provided the same answer for two questions. Why is it flagged as conflicting?
The AI flags FAQs as conflicting when they are very similar. You can keep both, but it's better to keep just one. This makes it easier to manage. If you need to update the answer later, you won't have to change it in multiple places.
Last updated