16:13
The human insights missing from big data | Tricia Wang
TED
Overview
This video explores the limitations of relying solely on big data for decision-making, arguing that it often misses crucial human insights. The speaker, a technology ethnographer, contrasts the ancient practice of consulting oracles with modern big data analytics, highlighting how both can be flawed without human interpretation. Using examples like Nokia's downfall and Netflix's success, the video advocates for integrating 'thick data'—qualitative, human-centered information—with big data to achieve more complete understanding and better outcomes, especially in complex human systems.
How was this?
Save this permanently with flashcards, quizzes, and AI chat
Chapters
- Historically, humans have sought certainty in decision-making by consulting oracles for predictions about the future.
- Modern society relies on big data, advanced algorithms, and AI as a new form of 'oracle' to predict outcomes and optimize processes.
- Despite massive investment, big data projects often yield surprisingly low returns, with many failing to improve decisions or foster innovation.
Understanding the historical desire for prediction and the modern shift to big data sets the stage for why a new approach is needed to make data truly valuable.
Ancient Greeks consulting an oracle for life-altering decisions like marriage or embarking on voyages.
- The speaker's ethnographic research with low-income populations in China revealed a strong desire for smartphones, contrary to Nokia's quantitative data.
- Nokia dismissed the qualitative findings because they didn't align with their existing big data models, which were designed to optimize current business, not emerging trends.
- Nokia's failure to incorporate these human insights led to a significant decline in their market share, illustrating the cost of missing emergent human dynamics.
This example demonstrates the critical risk of ignoring qualitative human behavior in favor of quantitative data, especially when predicting future market shifts.
Ethnographic work in China revealed that even low-income individuals were prioritizing buying 'shanzhai' (knock-off) smartphones, a trend Nokia's data did not capture.
- Big data excels in contained, predictable systems but struggles with the dynamic and unpredictable nature of human behavior.
- The 'quantification bias' is the unconscious tendency to overvalue measurable data, leading to the dismissal of immeasurable but crucial human factors.
- Over-reliance on quantifiable data can create an illusion of understanding while increasing the risk of missing critical, unpredictable 'tornado' events.
- Quantifying can be comforting and addictive, making it difficult to accept data that doesn't fit numerical models.
Recognizing the quantification bias is essential for avoiding the trap of believing we have all the answers when we only have a partial, measurable view.
Executives focusing solely on numbers from a big data system and ignoring qualitative evidence presented to them.
- The ancient oracle's effectiveness relied on 'temple guides' who gathered qualitative, ethnographic data (emotions, context, motivations) to interpret the oracle's predictions.
- 'Thick data' refers to rich, qualitative information from humans—stories, emotions, interactions—that cannot be easily quantified but provides deep meaning.
- Thick data complements big data by providing context, rescuing the loss that occurs when making data usable for machines, and leveraging human intelligence.
Introducing thick data as a necessary partner to big data provides a framework for incorporating human understanding into data-driven decision-making.
Temple guides in ancient Greece asking follow-up questions to understand the inquisitors' emotional state and motivations before interpreting the oracle's babblings.
- Netflix's quantitative data showed incremental improvements to its recommendation algorithm, but missed a larger user behavior.
- Hiring an ethnographer revealed the 'thick data' insight that users loved to 'binge-watch' shows and felt no guilt about it.
- By integrating this human insight with their big data, Netflix redesigned their platform to encourage binge-watching, leading to significant business transformation and growth.
This case study illustrates the power of combining big and thick data to uncover transformative insights that quantitative data alone cannot reveal.
Netflix changing its recommendation strategy from suggesting similar shows to offering more episodes of the same show to facilitate binge-watching.
- Integrating thick data is not just about business success; it's crucial for ethical decision-making, especially for marginalized communities.
- Big data in predictive policing and sentencing can reinforce existing biases if not tempered with human context and understanding.
- As automation increases, the quantification bias poses risks across various aspects of life, from healthcare to employment.
This highlights the broader societal implications of the quantification bias and the urgent need for a balanced approach to data to ensure fairness and avoid harm.
Predictive policing algorithms reinforcing existing biases against certain communities due to a lack of qualitative context in the data.
Key takeaways
- Big data provides scale and efficiency, but often lacks the depth of understanding that comes from human context.
- Qualitative 'thick data' (stories, emotions, interactions) is essential for interpreting and contextualizing quantitative 'big data'.
- The 'quantification bias' leads us to undervalue immeasurable human insights, which can be critical for predicting future trends and avoiding errors.
- Ethnographic research and qualitative methods are vital 'temple guides' for the 'oracles' of big data systems.
- Integrating big and thick data leads to more complete understanding, better decision-making, and potential business transformation.
- Ignoring human insights in data analysis can perpetuate societal biases and lead to harmful outcomes, especially for vulnerable populations.
- The future of effective decision-making lies in the synergy between machine intelligence (big data) and human intelligence (thick data).
Key terms
Big DataThick DataQuantification BiasTechnology EthnographerQualitative DataQuantitative DataEmergent Human DynamicsPredictive PolicingInformal EconomyShanzhai
Test your understanding
- What is the 'quantification bias' and how does it hinder effective decision-making?
- How does 'thick data' complement 'big data' in providing a more complete understanding of a situation?
- Why was Nokia's reliance solely on quantitative data a critical mistake, according to the speaker?
- What lessons can be learned from the ancient Greek oracle's method of prediction that apply to modern big data analysis?
- How can integrating big and thick data lead to transformative business outcomes, as exemplified by Netflix?