Conversational Agents Flashcards
Conversational (Interface)
Conversational interfaces are
interfaces that enable people to
interact with smart devices using
conversational spoken language.
* Voice (e.g. Alexa)
* Text (e.g. ChatGPT)
Agent
Agents are autonomous, active
computer processes that possess
some ability to communicate with
people and/or other agents and to
adapt their behavior.
Conversational Agents
Conversational agents are
software-based systems designed
to interact with users using natural
language
Main Components of Conversational Agents
Main components
◼ Speech recognition: converts the speech input
into a string of words
◼ Natural language understanding: interpret the
user’s input and to extract a representation of its
meaning
◼ Dialog management: tracks the state and flow of
the conversation and controls how the system
responds to the user’s input
◼ Natural language generation: Formulate a
response in natural language
◼ Text-to-speech synthesis: converts the response
string into speech output
Classification of conversational Agents
Classification of conversational agents based on two dimensions
◼ 1) Mode of communication: text vs. voice
◼ 2) Context: open-domain vs. task-oriented
Shortcomings of conversational Agents
Ambiguous nature of human language
→ natural language processing is difficult
Understanding context, sarcasm, etc.
Not everything can be transferred via voice
→ Sometimes visual information would be helpful
Socia Cue Definition
A social cue is any design feature
that presents a source of
information and triggers a social
response from the user towards
the chatbot
Socail Response Theory
Social cues from computers trigger mindless
responses from humans, no matter how rudimentary those cues are
Cue -> Mindless Behavior -> Social Responses
Example for Socail Cues
Verbal: Style: Formality
Visual: Kinesics: Gestures
Auditory: Voice Qualities: Voice Gender
Invisible: Chronemics: Response Time
Design Principle 1
DP1: Provide the CA with advanced natural language processing capabilities and business integration in order for users to communicate their service request in a natural way and receive a high-quality response.
Based on cooperative principle
Design Principle 2
Provide the CA with conversational abilities to deliver messages characterized by situation-dependent length and segmentation in order for users to receive the right amount of information.
Based on cooperative principle
Design Principle 3
Provide the CA with perspicuous and flexible conversation flows towards a specific goal combined with effective clarification, confirmation, and error-handling strategies in order for users to efficiently achieve their goals despite potential misunderstandings.
Based on cooperative principle
Design Principle 4
Provide the CA with social cues that correspond to key service agent characteristics, fit its context, and neither over- nor underplay its capabilities in order for users to better understand what they can expect from it.
Based on Social Response Theory