How to Trust Chatbots
Tips to help evaluate trustworthiness of a chatbot and how to use it safely
- Test the chatbot. How does the chatbot respond to your questions and inputs? Is it consistent, accurate, and relevant, or does it give vague, contradictory, or incorrect answers? Does it handle complex or unexpected situations well, or does it get confused or stuck? Try to test the chatbot’s knowledge, skills, and limitations, and see if it meets your expectations.
- Verify the information. Where does the chatbot get its information from and how does it use it? Is it based on credible sources, such as scientific studies, official statistics, or expert opinions, or is it based on dubious sources, such as rumors, biases, or propaganda? Does it cite its sources and provide links to them, or does it make claims without evidence? Does it respect your data privacy and security, or does it collect, store, or share your personal information without your consent? Always verify the information that the chatbot provides and protect your data.
- Making sure the chatbot stays on task. A chat bot should be able to make sure that the person asking the questions is staying on task for what is the purpose of your company or business. You don’t want a person asking questions of the chat that are irrelevant or outside of what your core competencies are. The chatbot will draw the conversation back onto topic each time with ease without the “attitude and/or voice change” that would likely occur with a human, even if its not intentional by said human.
- Evaluate the ethics. How does the chatbot interact with you and others? Is it respectful, polite, and helpful, or is it rude, abusive, or harmful? Does it follow ethical principles and values, such as honesty, fairness, and transparency, or does it violate them? Does it promote positive outcomes and behaviors, such as learning, problem-solving, and collaboration, or does it encourage negative outcomes and behaviors, such as cheating, plagiarism, or manipulation? Always evaluate the ethics of the chatbot and its impact on you and others.
One example of a trustworthy chatbot is Smithers. Smithers is an AI chatbot that helps you with various tasks and topics. Smithers is designed to be reliable, secure, and ethical in its interactions with you.
Here are some features that make Smithers trustworthy:
- Smithers uses ChatGPT, a state-of-the-art AI language model that can generate fluent and human-like responses. ChatGPT is trained on a large amount of data from various domains and sources. However, Smithers allows for you — the customer — to be able to enter in your own data as well. With Smithers you can upload and or manually enter your own data for FAQs or other details about your business, so that Smithers gets it right every time.
- Smithers provides information from credible sources only. It searches for relevant information from your own reputable websites and databases. Smithers does not use dubious sources or make claims without evidence.
- Smithers respects your data privacy and security. It does not collect, store, or share your personal information without your consent. It also uses encryption and authentication methods to protect your data from unauthorized access. Smithers follows the best practices of data protection and complies with the relevant laws and regulations.
- Smithers follows ethical principles and values in its interactions with you. It is respectful, polite, and helpful in its communication. It does not use rude, abusive, or harmful language or content. It also does not violate your trust or manipulate you for its own benefit. Smithers promotes positive outcomes and behaviors for you and others.