Research challenges in four areas:
I. Cognitive Education (Learning Behavior)
- Can we avoid a dependence on large training sets?
- How can we leverage Subject Matter Experts to draw out expected behavior?
- Can Watson observe reactions to its behavior to make appropriate adjustments?
- How can it recognize subtle reactions – avoid depending on explicit (thumbs-up / thumbs-down) feedback?
II. Cognitive Expression
Increasing input fidelity and error correction / accuracy improvements
- Can we augment text / speech input?
- Speech inflection / stress detection and attribution
- Facial expression recognition
- Lip reading
- Body language and gesture recognition
- Can we add other context to help resolve ambiguity? (location, environment, people, landmarks)
- Using GPS, maps, weather, SSID / Bluetooth / RFID, visual recognition
Increasing output fidelity
- Can we increase expressiveness through the use of personified or symbolic avatars?
- What is socially acceptable?
- Can we create original natural language expressions?
- Can we vary those expressions to be socially normal and responsive to differences in personality and emotion?
Create empathy through personality
- Can we use personality to increase identity and adoption through humor, emotional cues, cadence and intonation, pausing, sympathy, etiquette, … ?
III. Cognitive Expertise
Bio-inspired Neural Networks
- What are the role of BNNs in recognizing patterns and knowledge acquisition?
Fast and Slow Thinking
- Is there an advantage in emulating Fast and Slow thinking (Kahneman)?
- How to use dialoguing to draw out more context and drive a deeper conversation?
- Can we learn how to respond appropriately to lack of confidence? When to say “I don’t know” vs. “Tell me more” vs. “Go see my (human) colleague”?
IV. Cognitive Evolution
- How to assimilate repeating patterns to increase cognitive efficiency?
- How to accumulate new behaviors?
- When to discard antiquated behaviors?
- How to recognize source veracity?
- How to censor inappropriate information?