Mixed-Initiative Context Filtering and Group Selection for Improving Ubiquitous Help Systems

Nasim Mahmud, Kris Luyten, Karin ConinxIsami 2012 · DOI

TL;DR

This paper introduces the Ubiquitous Help System (UHS), a mobile system designed to help users quickly find relevant social contacts for assistance. UHS automatically extracts and prioritizes contextual information from a user's query and sensor data, then intelligently selects a group of potential helpers from their social network. This aims to reduce the "social cost" of seeking help by making the process more efficient and reliable.

What Problem Does This Paper Address?

In our increasingly mobile world, people often need help or information but struggle to find the right person to ask, especially when they're away from their usual social circles. Traditional methods of asking for help, like posting on social media, can be time-consuming and lack crucial context. This paper tackles the challenge of efficiently connecting a help-seeker with the most appropriate social contacts by automatically understanding the context of their request and the availability of their network. It addresses the difficulty of conveying rich contextual information, especially without relying on speech, and the problem of identifying suitable helpers from a potentially large social network.

Methods

The core of UHS is its mixed-initiative approach, meaning it combines automated processing with user control.

  • Context Selection and Prioritization:
    • The system takes a user's voice query, converts it to text, and then analyzes it for meaning using a semantic dictionary like WordNet.
  • It identifies keywords and their synonyms to understand the core of the question.
  • Crucially, it also leverages sensor data from the mobile device (like location, time, and potentially activity inferred from speed) to enrich the context.
  • This contextual information is then prioritized based on its relevance to the query, creating a ranked list of context elements to be sent with the question.
  • Group Selection:
    • Once the relevant context is identified, the system selects a group of potential help providers from the user's social network.
  • This selection considers factors like the user's current location, destination, and the availability status of their friends (e.g., from their to-do list or system settings).
  • An "UsersAffinity" function dynamically assesses a user's connection to the current situation or destination.
  • The algorithm then refines the list of available friends, prioritizing those who are both socially "nearby" (frequently interact with the user) and relevant to the selected context.
  • The system allows users to manually intervene and edit the suggested group, offering a balance between automation and user control.

Datasets

Not explicitly discussed in the paper. The system relies on sensor data from typical Android phones (GPS, camera, voice recorder) and uses the WordNet semantic dictionary. It also implicitly uses the user's social network and potentially their to-do list data.

Key Insights and Ideas

  • Context is King (and Queen): The paper strongly emphasizes that simply asking a question isn't enough; the context surrounding the question is vital for getting a useful answer. This includes not just what the user is asking about, but also where they are, when they're asking, and their current activity.
  • Social Networks as a Resource: Instead of relying solely on public knowledge bases or search engines, UHS leverages the user's existing social network as a primary source of help, recognizing that friends and colleagues often provide more reliable and tailored advice.
  • Mixed-Initiative for Usability: The "mixed-initiative" approach is key. It automates much of the complex context gathering and helper selection, but crucially allows the user to review and adjust the suggestions. This prevents the system from making incorrect assumptions and ensures user satisfaction.
  • "Social Cost" Reduction: By automating the process of identifying who to ask and what information to provide, UHS aims to significantly reduce the effort and time a user expends when seeking help, thereby lowering the "social cost" of asking.

Key Results

The paper presents algorithms for context selection and group selection. While specific quantitative results (e.g., accuracy percentages, speedup metrics) are not detailed in this excerpt, the paper demonstrates the feasibility of:

  • Automatically extracting and prioritizing contextual information from voice queries and sensor data.
  • Selecting a relevant group of potential help providers from a user's social network based on context and availability.
  • The mixed-initiative interface allows users to intervene and refine the system's suggestions, indicating a qualitative success in user control.

Limitations and Future Work

  • Understanding Internal Context: The authors acknowledge that fully understanding a user's internal context (e.g., their precise mental state or nuanced intentions) is difficult to measure with current sensors.
  • Reliability and Comfort: While the system aims to improve reliability by selecting relevant helpers, the paper notes that the "social cost of seeking help" is still a factor, and the system's effectiveness relies on the willingness and expertise of the social network.
  • Future Work: The paper suggests that the semiautomatic approach, allowing user intervention, is beneficial. Future work would likely involve further refining the algorithms for context prioritization and group selection, potentially incorporating more sophisticated models of social relationships and user intent. The paper also implies the need for more extensive user studies to validate the system's effectiveness in real-world scenarios.

Summary updated

BibTeX
@inproceedings{mahmud2012mixed,
  author = {Nasim Mahmud and Kris Luyten and Karin Coninx},
  title = {{Mixed-Initiative Context Filtering and Group Selection for Improving Ubiquitous Help Systems}},
  booktitle = {Isami},
  year = {2012},
  doi = {10.1007/978-3-642-28783-1\_19},
}
Figure