AI summary by TwinMind
Anne Scherer: Why we're more honest with machines than people | TED
TED
11m 18s
Barbecue Anecdote and Social Rules Introduction
- Researcher's barbecue story: adult mentioned lack of sex with wife → awkward silence
- 6-year-old boy blurted out parents have lots of sex → laughter, conversation continued
- Both broke social rule: never discuss sex, money, politics at dinner table
- Adults expected to know/follow rules → sanctioned with ignorance when broken
- Children given pass due to naive understanding of social manners
Social Norms and Their Evolution
- No official rulebook for socially appropriate behaviors
- Social norms = unwritten codes of conduct that change over time
- Example: handshaking pre-COVID considered polite → post-COVID frowned upon
- Social rules learned through rewards and punishments
- Humans seek social approval, avoid disapproval
- Present ourselves as smart, successful, sporty, active, creative, empathic
Social Media Impact on Self-Presentation
- Social media skyrocketed strive for social approval
- Increased need for self-presentation and perfection
- Flip side: constant fear of disapproval when can't meet expectations
- Examples: adults with incontinence, drug addiction, returning soldiers
USC Army Research Case Study
- Traditional process: soldiers interviewed by human healthcare professionals post-combat
- Problem: soldiers hardly reported any problems upon return
- Suspected many didn't dare share problems openly
- Soldiers trained to be strong/brave, not show weaknesses
- Difficult to admit health problems, sleep troubles, nightmares
SimSensei Virtual Interviewer Solution
- USC team developed virtual interviewer called SimSensei
- Digital avatar with humanlike appearance
- Conducted natural conversations with clients
- Results: soldiers reported MORE health problems (nightmares, sleep troubles) with SimSensei
- Machines remove "social" from interaction → people open up more easily
Researcher's Chatbot Studies
- Developed series of chatbots with colleagues
- Two types designed:
- Humanlike: social, used speech disfluencies ("ohos", "ahas", "hmms")
- Machine-like: functional, lacked social cues, stuck to talking points
- Questions gradually became more personal/delicate
- Hypothesis: people apply more social rules with humanlike bots
Research Findings on Social Evaluation
- Participants with humanlike bots showed more concern about social evaluation
- Result: gave more socially desirable responses
- Specific example - prior sex partners question:
- Men with humanlike bot: reported significantly MORE partners
- Women with humanlike bot: reported significantly LESS partners
- Compared to those interacting with mechanistic bot
Key Implications and Takeaways
- Gender differences in social desirability: men want more partners, women want fewer
- Consumer psychology insight: people open up more with obviously machine-like interfaces
- Current trend: massive investment making machines indistinguishable from humans
- Sometimes "letting machine be machine" is beneficial
- Machine interactions criticized for making kids rude/impolite
- Upside: machines provide break from constant self-presentation pressure
- Don't need to worry about being judged for personal details
- Machines remind us of good conversation partner trait: being nonjudgmental
Call to Action
- Encouragement to be less judgmental in social situations
- When others openly share thoughts/feelings/problems, respond like machines do
- Be nonjudgmental conversation partners
