The human insights missing from big data | Tricia Wang

The Role of Oracles in Decision-Making

Ancient Practices of Seeking Guidance

• In ancient Greece, individuals from various backgrounds consulted oracles for significant life decisions, such as marriage and military actions.

• The oracle would enter a trance state to provide predictions after a period of time, reflecting the human desire for certainty in decision-making.

The Modern Oracle: Big Data

• Today, big data serves as our modern oracle, with technologies like Watson and deep learning answering complex questions about logistics and health.

• People seek predictive insights to alleviate fears about uncertain futures, similar to how ancient societies sought prophecies.

Challenges in Utilizing Big Data

Industry Insights on Big Data Investments

• Despite being a $122 billion industry, over 73% of big data projects fail to yield profits due to ineffective implementation.

• Executives report that investments in big data systems do not lead to improved decision-making or innovative ideas among employees.

Personal Experience with Nokia

• As a technology ethnographer, the speaker shares experiences from working with Nokia during its peak years and observing low-income users' technology adoption patterns.

• The speaker's qualitative research revealed an emerging demand for smartphones among low-income Chinese consumers despite skepticism from larger corporations like Nokia.

The Disconnect Between Data and Insight

Misalignment Between Qualitative Insights and Quantitative Data

• Nokia dismissed the speaker's insights based on limited qualitative data compared to their vast quantitative datasets, leading them to overlook market trends.

Understanding the Complexity of Human Systems and Data Integration

The Challenge of Quantifying Dynamic Systems

• Many systems, such as electricity grids or genetic codes, can be quantified effectively when they are contained. However, human-involved systems are dynamic and unpredictable, making them harder to model accurately.

• This unpredictability leads to a cycle where new unknown factors emerge after predictions about human behavior are made, highlighting the limitations of relying solely on big data.

• The "quantification bias" refers to the unconscious preference for measurable data over immeasurable insights, which can skew our understanding of complex situations.

• In workplaces, this bias manifests as an obsession with numerical metrics that blinds individuals and organizations to qualitative evidence right in front of them.

• While quantifying data is satisfying and provides comfort (e.g., through spreadsheets), it can lead to addictive behaviors that overlook critical non-numerical information.

The Dangers of Oversimplification

• Relying too heavily on quantifiable data can result in "silver-bullet thinking," where organizations seek simple solutions for complex problems, risking blindness to significant issues.

• Historical context from ancient Greece illustrates this point; the oracle at Delphi was influenced by environmental factors that altered her state but required guidance from temple guides for meaningful interpretation.

Integrating Big Data with Thick Data

• Just as the oracle needed support from guides, big data systems require qualitative insights from ethnographers and user researchers who gather what is termed "thick data."

• Thick data encompasses rich narratives—stories and emotions—that provide depth beyond mere numbers. It helps identify gaps in existing models by grounding business questions in human experiences.

• Combining big data's scalability with thick data's contextual richness creates a more comprehensive understanding. This integration allows exploration beyond collected datasets into uncharted inquiries about underlying reasons for observed phenomena.

Case Study: Netflix's Transformation Through Ethnography

• Netflix sought to enhance its recommendation algorithm through quantitative means but found only incremental improvements despite offering a $1 million prize for better algorithms.

Netflix's Strategy for Binge-Watching

Redesigning Viewer Experience

• Netflix implemented a strategy to enhance binge-watching by offering more of the same show rather than diversifying genres or shows. This approach simplifies viewer choices and encourages prolonged viewing sessions.

• The integration of big data and thick data has not only improved Netflix's business model but also transformed media consumption patterns, leading to significant stock projections.

Implications of Big Data

• The use of thick data insights in algorithms can have critical implications, particularly for marginalized communities, where predictive policing practices may reinforce existing biases.

• Examples include police departments using big data for bond amounts and sentencing recommendations, which can perpetuate systemic inequalities.

Consequences of Misuse

• The NSA's Skynet machine learning algorithm is cited as having potentially contributed to civilian casualties in Pakistan due to misinterpretation of cellular metadata, highlighting the dangers of flawed data interpretation.

Future Considerations

• As automation becomes prevalent across various sectors (automobiles, health insurance, employment), individuals will increasingly face challenges related to quantification bias.