In 1943, psychologist Abraham Maslow analyzed human motivation as a hierarchy of needs: physiological needs, safety, love and belonging, esteem, and self-actualization. But Maslow could hardly have imagined the digital age in which we live today. Our modern existence in an ocean of petabytes calls for its own hierarchy of needs — a framework for how we interact with and benefit from data.
I propose that we break these needs down into three levels: information, attention, and trust.
Much as our physiological needs constitute our most basic requirements for survival, information is the foundation for our digital existence. Fortunately, in today’s world of nearly-free digital distribution, information is everywhere — and it’s free!
Well, not exactly. Even though many of us complain about information overload, much of the world’s information hasn’t been digitized — let alone optimized for human and machine consumption. And often information only exists as latent structure in data, waiting to be discovered through analysis.
We are still in the early days of making information digitally accessible, from Google scanning library books to LinkedIn and Facebook building online representations of our professional and social networks. Indeed, data scientists like my team at LinkedIn spend most of our time converting massive volumes of data into useful information — not just for people to consume directly, but also to power other analyses and products.
In the digital age, information is not only our food and water, but also the fuel for our machines.
But what about information overload? Don’t we have too much information already?
We certainly have what Herbert Simon predicted would be the consequence of a wealth of information: a poverty of attention, and a need to allocate that attention efficiently. Or, asClay Shirky puts it, our problem is not so much information overload as filter failure. We need better ways to route and allocate our attention.
Technology’s answers to this challenge are relevance and personalization. Search engines use relevance ranking algorithms to guide users toward the information they want or need. Recommender systems similarly push information that is optimized for a relevance metric, such as click-through rate. Meanwhile, personalization represents and leverages differences among users to better match them with information.
But ultimately we have to optimize the allocation of attention by explicitly modeling attention scarcity. Influence measures like TunkRank recognize human attention as a finite quantity.Attention bond mechanisms fight spam by enabling recipients to place a monetary value on their attention. Treating attention as an economic good is essential if we are to allocate attention effectively, let alone optimally.
If information provides our digital sustenance, then attention determines how we assemble a nutritious and delicious diet.
At the top of our pyramid of digital needs is trust. Even if all the information we need is available, discoverable, and relevant, we still need to know whether we can rely on it. And, since we are often producers of information ourselves, we have an interest in establishing our own trustworthiness as sources.
The nature of trust deserves more than a few paragraphs. But for the purpose of this framework, we can think of trust in terms of two dimensions: authority and sincerity.
Authority is the degree to which the source of information possesses the requisite knowledge or expertise to be reliable, i.e. the source’s ability to deliver objective truth. Sincerity is whether the information provided in good faith, i.e., the source’s desire to deliver subjective truth. Both are crucial — and sincerity is especially important to consumers of subjective information, such as product reviews.
We mostly rely on two mechanisms for building trust: history and networks. History is the assumption that past trustworthiness is predictive of future trustworthiness, with the twist that we heavily penalize even a single breach of trust. Networks propagate trust among people and other information sources — a crucial mechanism that enables trust to operate at scale, but also makes the network acutely vulnerabile to deception.
Completing the sustenance metaphor, trust certifies the safety and integrity of the information we consume.
A Lifetime of Research Challenges
Information, attention, trust: these are our digital needs. And, like Maslow’s needs, they form a hierarchy when each level depends the previous one.
The science and technology to satisfy these needs are in their infancy, and many computer scientists and social scientists aim to advance the state of the art. On a personal note, this framework neatly summarizes the scope of my professional interests. I cannot imagine more interesting challenges than those relating to information, attention, and trust. I hope that my life’s work will help address some of those challenges.