AI summary by TwinMind
How Targeted Ads Might Just Save Your Life | Sandersan Onie | TED
TED
10m 36s
Personal Story and Problem Context
- Speaker diagnosed with depression in 2015, experienced multiple intense suicidal episodes
- Nearly ended life but was saved when someone reached out
- Australia: suicide is leading cause of death for ages 15-44
- For every suicide death, 25-30 times more attempts occur
- Less than half of people who attempt suicide reach out for help prior
- Need to meet people where they are rather than wait for them to seek help
Research Foundation
- Most people who attempt/think of suicide don't actually want to die
- They're looking for any reason to continue living
- Any interruption (tap on shoulder, message, internet pop-up) could save lives
- Current hotlines/emergency services require individual to take first step
- Challenge: how to know what person is thinking/feeling if not speaking up
Search Behavior as Intervention Point
- People search things online they'd never ask another person
- Study: individuals who recently attempted suicide had searched suicide-related keywords
- Searched suicide means and methodology
- Searched terms about hopelessness and loneliness
- Research from Japan, US, Taiwan shows volume of suicide-related searches correlates with national suicide rates
- Australians spend 40 hours/week on internet, 93% starts on search page
- Current intervention: hotline banners appear for suicide searches
- Not everyone wants to call hotlines
- Many countries lack sustainable hotlines
- Certain keyword combinations don't trigger banners, may show harmful information
Proposed Solution: Algorithmic Intervention
- Use same algorithms tech companies use for ad targeting
- Identify individuals thinking of suicide using age, gender, language, geography
- Reach out in deeply personal and engaging way
- Example: Amy (22-year-old Sydney girl)
- Living with housemates, cheerful demeanor but suicidal thoughts
- Family unhelpful, turns to search for suicide methods
- Algorithm triggers, shows ad leading to landing page
- Page designed to lead away from suicide attempt
- Features stories of others who've been in similar situation
- Provides multiple help-seeking options
Development Process and Targeting
- Collaborated with people who've thought of suicide, attempted suicide, lost someone to suicide
- Developed keyword lists people likely to search
- Created ad content and landing page design
- Landing pages must be maximally engaging (may be minutes from attempt)
- Extensive targeting by age, language, gender, geography for personalization
Age-Based Customization Examples
- Teens: prefer calming/distracting resources, breathing exercises
- Older generation (40s-50s): prefer direct, no-nonsense, straight-to-point language
- Robert example (55-year-old rural Queensland man)
- Economic stress, feels can't talk to anyone as breadwinner
- Searches life insurance policy suicide clauses
- Shown direct problem-solving language appropriate for his demographic
- Welcome videos feature people from same age group
Trial Results in Australia
- 19-day campaign reached over 120,000 people searching distress/suicide keywords
- Industry standard engagement: approximately 3%
- Achieved engagement rate: over 22% (7x industry average)
- Indicates real need for this intervention type
Cultural and Language Adaptations
- Developing pages for Aboriginal and Torres Strait Islanders
- Creating Mandarin-speaking population resources
- Indonesia implementation:
- Mental health stigma severe (families may cage mentally ill members)
- Over 21 unique regional dialects
- Geographic targeting shows different dialect pages by location
- Samarang (Central Java): Bahasa Java pages
- Palembang: Bahasa Palembang pages
- Currently used in Kidul Mountain (high suicide rate area)
- Positive reception because regional language "speaks to hearts" vs national language "speaks to minds"
Broader Vision and Limitations
- Entire tech industry has technology, data, means to reach people in need
- Currently only used to tailor ads and sell products
- Cannot replace relationships and communities
- Since most don't seek help prior to attempts, this fills important puzzle piece
- Goal: ensure every person in dark time meets "outstretched hand"
- Intervention methods: tap on shoulder, text message, internet ad
