The inclusion of unsolicited content in a human-to-computer conversation can improve the user experience. It makes automated assistants more proactive and helpful. These systems provide information that may be of interest to the user, even when not explicitly requested. This leads to more engaging and productive interactions.
Why Add Unsolicited Content in A Human to Computer Dialog?
Here are some key reasons why unsolicited content can be beneficial in human-to-computer dialogues:
Automated assistants that only respond to specific user requests tend to be reactive. This reactivity may limit their usefulness and engagement potential. By incorporating proactive suggestions or information, these assistants can maintain an ongoing conversation, making them feel more responsive and intelligent.
This can help users stay engaged and discover functionalities they may not have considered.
Anticipatory Assistance: After completing its tasks and receiving no new input, the automated assistant may offer helpful, unsolicited content based on contextual knowledge, preferences, or past interactions.
After setting an alarm, the assistant might ask, “Would you like a reminder about your upcoming meeting this afternoon?”
By anticipating the user’s needs, the assistant provides convenience without requiring the user to think of every potential request.
Discovery of Features: Users often don’t know everything an assistant can do. By proactively offering additional information, tips, or services, the automated assistant can highlight its capabilities and show users how they might benefit.
For example, after assisting with a calendar appointment, the assistant might say, “By the way, I can help you find nearby restaurants for your lunch break if you’d like.”
Contextual Personalization: Adding proactive content provides an opportunity for personalization. If the assistant has access to user data such as preferences or routines, it can offer suggestions tailored to the user’s habits or interests.
For instance, if a user greets the assistant in the morning, the assistant might provide relevant news updates, the weather forecast, or reminders based on their calendar.
This enhances the user’s perception of the assistant as being more intelligent and attentive to their needs.
Reducing Friction in User Interaction: Many users hesitate to initiate interactions beyond basic requests, either because they are unsure of the assistant’s capabilities or because of a lack of habit.
Providing proactive content helps reduce this friction, encouraging the user to explore and engage more deeply with the assistant.
It can also make the experience feel less formal and more conversational, as the assistant appears more like a human conversation partner.
Increasing Productivity: By providing additional, potentially useful information when the assistant is idle, users can benefit from quicker access to services or suggestions that they otherwise might not have thought to ask for.
For example, after asking about the weather, the assistant might proactively ask, “Would you like me to set an umbrella reminder if it looks like rain later?”
This kind of proactive offer can save the user time and effort.
Advantages of Inserting Unsolicited Content Into a Human-to-Computer Dialog Session
The advantages of inserting unsolicited content into a human-to-computer dialogue session can be summarized as follows:
- More Lifelike Interaction: Automated assistants that incorporate unsolicited content can seem more “lifelike” or human-like. This enhanced naturalness in interaction can lead to increased user engagement.
- Personalized Content Delivery: The automated assistant selects unsolicited content based on user-specific characteristics like location, travel plans, or interests to provide highly relevant information, making the experience more engaging and useful.
- Reduced Input Effort: Receiving unsolicited content can be very helpful for users with limited abilities to provide input, such as when driving or due to physical limitations. It eliminates the need for active requests, simplifying interaction.
- Discovering Useful Information: Users may receive helpful information that they might not have thought to ask for. This proactive approach helps users discover useful content and new functionalities of the assistant.
- Efficiency in Information Retrieval: Unsolicited content can reduce the need for users to make additional requests, saving time and conserving computing resources that would otherwise be needed to handle these requests. It streamlines the process of information retrieval, making interactions more efficient.
Ways that A Computer May Provide Unsolicited Content
The patent describes a variety of signals that could be used by an automated assistant to provide unsolicited content to the user. These signals help the assistant select information or actions that may be of interest to the user based on contextual clues. Some of these signals include:
- Past Dialogues: Previous interactions between the user and the assistant can guide unsolicited content. For example, if a user has searched for flights but hasn’t booked tickets. The assistant might ask relevant questions or suggest deals related to the destination.
- User Location: The user’s current location could trigger suggestions such as nearby restaurant specials, reminders of calendar entries, or events relevant to the location.
- Calendar Events and Reminders: The assistant might proactively remind the user of upcoming calendar entries (e.g., anniversaries or appointments). For example, it could suggest packing for an upcoming flight or remind the user to perform an online check-in.
- Search and Browsing History: Based on the user’s search or browsing activity, the assistant could provide related information. For instance, if the user has been researching a particular sports team, the assistant might bring up the latest scores.
- Topics of Interest: The assistant could suggest content based on the user’s known interests, such as favorite sports teams, hobbies, or events.
- Email and Documents: Emails or other documents could prompt the assistant to remind the user of upcoming events or follow up on specific content. For example, an email invitation could lead the assistant to remind the user of the event’s date.
- Application States: The assistant could provide suggestions based on the user’s current application use. For example, it might remind the user to close applications that are using device resources, notify of application updates, or highlight new features that may be useful to the user.
- New Features: If the assistant has new features, it could proactively inform the user. For example, “While you were away, I learned how to call a taxi. Just let me know if you need one.”
- Weather Information: The assistant could use weather data to offer relevant suggestions. For instance, on a sunny day, it might suggest restaurants with outdoor seating options.
The signals help the automated assistant personalize the user experience by providing relevant content or reminders based on the user’s current context or interests. No explicit user prompts are needed.
Conclusion
It’s interesting to observe how Google’s patents explore the nuances of natural language interaction between humans and automated assistants. This latest approach emphasizes proactively inserting relevant, unsolicited content during a conversation. It aims to make these assistants appear more human-like and helpful, especially when users change topics or need assistance without being prompted.
Other patents have indeed focused on various aspects of human-computer dialog beyond topic shifts or proactive content. I plan to delve into those soon as well.
Next time you interact with your assistant, pay attention to any unexpected suggestions or information it offers. It might be using some of these new methods!
The Unsolicited Content in Human to Computer Dialog was based on the following patent:
Proactive incorporation of unsolicited content into human-to-computer dialogs
Inventors: Ibrahim Badr, Zaheed Sabur, Vladimir Vuskovic, Adrian Zumbrunnen, and Lucas Mirelmann
Assignee: Google LLC
US Patent: 11,232,792
Granted: January 25, 2022
Filed: March 25, 2020
