Juji's mission is to discover the world’s people by their individuality, and to enable individualization at scale. Our notion of individualization includes two aspects. One is self individuation, a concept proposed by the renowned Psychologist Carl Jung: one should be aware of and realize one's true self. The other is for individuals to obtain hyper-personalized experiences that suit their true nature, including their personality, physical and emotional needs, values and beliefs, and taste and style.
As the first step toward fulfilling our mission, we empower our users, individuals and businesses alike, to automatically derive deep insights about people of interest from opt-in or public data (e.g., public blogs and web pages). Such insights characterize a person's or a group's individuality from multiple aspects, encompassing their professional and personal lives. Our users can then use the derived insights to develop and improve their own brand personality, and create individualized engagements at scale (e.g., interacting with hundreds or thousands of customers in an personalized way).
Juji is built upon many years of scientific research and studies that intersect multiple disciplines, including Psychology, Linguistics, Sociology, Behaviorial Economics, and Computer Science. Its core idea is rooted in three premises:
The first premise is formed based on extensive, empirical research in Psychology, Humanity Studies, and Behavorial Research. The findings show that one's individuality traits are multi-faceted and complex. For example, researchers have developed a number of models to characterize one's personality traits alone (e.g., [Johnson & Ostendorf 1993, Hofstee 1992, Saucer & Goldberg 1996]). More importantly, individuality traits are found to be powerful predictors of people's lives in many aspects.
In general, research shows that people are happier if they live lives that match their true nature and capabilities [Schacter et al. 2011]. Based on these studies, we focus on modeling individuality traits that matter—having significant effects on important life outcomes (e.g., occupational proficiency and well-being).
The second premise is grounded in a rich body of scientific discoveries that reveal a person's communication signals reflect one's individuality, such as personality, values, political orientation, cognitive style, and emotional traits. For example, in "Honest Signals", Pentland describes how one's unconscious body behavior signals a person's attitude and values [Pentland 2010]. Pennebaker, meanwhile, in "The Secret Life of Pronouns", shows that word uses, especially the use of functional words like pronouns, reflect a person's personality, thinking style, and emotional characteristics [Pennbaker 2011]. More recently, this line of research has been applied to analyzing people's extensive online communication activities, and correlating their activities with self-reported trait assessment. For example, Yarkoni correlates the word use in one's blogs with their Big 5 personality test results [Yarkoni 2010]. Similarly, Kern et al. examine the correlations between people's Facebook posts and their Big 5 personality traits [Kern et al. 2013], while Kosinski et al. report the relationships between one's Facebook likes and a variety of traits, including personality and political orientation [Kosinski et al. 2013]. Inspired by this line of research, we examine verbal and social interaction signals in one's communication activities, and explore the use of such signals to infer one's specific individuality traits in various contexts (e.g., workplace or personal relationship).
The third premise arises from the proliferation of user-genered data and recent advances in computing—the ability to process big data and derive insights from the data. In particular, there are numerous research efforts on analyzing user-generated data and infer various traits of users. For example, Golbeck and her team have used one's social activities on Twitter and Facebook to infer their Big 5 personality traits [Golbeck et al. 2010, Adali et al. 2012], while Shen and his colleagues have used one's email to infer their personality traits [Shen et al. 2013]. In addition to deriving personality traits, Pennacchiotti and Popsecu use one's tweets to derive a person's gender, age, and political orientation [Pennacchiotti 2011]. Previously, with our colleagues at IBM, we had also worked on using one's tweets to infer various individuality traits, including human basic values [Chen et al. 2013], needs [Yang and Li 2013], attitude [Huiji et al. 2014], and emotional style [Zhao et al. 2014]. Extending these works, which mostly focus on deriving basic and generic traits, such as Big 5 personality, we derive a set of composite traits that characterize a person in a specific context. For example, Juji infers a person's career DNA, which consists of a set of work-related traits, including drive and ambition, grit and flexibility, and work ethics.
One of the most asked questions about our work is
How accurate are the inferred traits?
One intuitive approach to validate the accuracy of derived traits is to compare the analysis results with that obtained from other means, such as self-reported psychometric test results and third-party assessment (e.g., asking a professional profiler or a friend to assess one's individuality traits). Many researchers have done such comparisons. For example, one of our own studies compared the computer-derived Big 5 personality traits, human basic values, and needs with those obtained from the person's corresponding psychometric test results. The results were quite encouraging: they showed significant correlation for over 80% of the test population [Gou et al. 2014]. More recently, Wu et al. also show that computers are better at predicting a person's Big 5 personality traits than humans do (e.g., the person's friends) [Wu et al. 2015].
Although comparing computer-inferred traits against human assessments might provide a certain level of validation, humans are known to be inconsistent and may not even be completely truthful in their tests or assessments. Therefore, we believe a better approach to assess the quality of the derived traits is to measure the effectiveness of the traits for predicting one's behavior in the real world. For example, if one or more traits are found to be a powerful predictor of one's real-world actions to entertain or shun away from new ideas, then these traits are the defacto "openness" trait of the person regardless the person's self assessment on his/her "openness". In other words, we want to ask
How useful are the inferred traits?
To assess the "usefulness" of inferred human traits, we examine the prediction power of the inferred traits on specific human behavior. Some of our previous research has done exactly so. For example, Mamhud et al. use a combination of derived people traits (e.g., Big 5 personality traits derived from tweets) to predict who is more likely to respond to strangers on open social media [Mamhud et al. 2013, Lee et al. 2014] or in an enterprise context [Luo et al. 2014]. The findings were very promising, since our algorithms were able to effectively predict who are more likely to respond to a stranger's request by their traits. Using the prediction model that we built, the machine significantly improved the response rates over a baseline by engaging with those who are more likely to respond. Overall, this line of research has shown great promises. And our goal is to advance this line of work by computationally discovering actionable human traits that affect one's real-world behavior so that they can be used to help guide and improve one's life outcomes.
Before starting Juji together, the two co-founders, Michelle and Huahai, had worked together for many years at IBM Research at Almaden, before becoming part of the IBM Watson Group. Although coming from different backgrounds, they both are passionate about developing state-of-the-art Cognitive Computing technologies that enable better computer-human symbiosis. Since at IBM, they have been deeply involved in such efforts, including the development of personality analytics technologies. They hoped to use the machine-inferred deep human insights to foster better human-computer interaction as well as human-human engagements. During their endeavor at IBM, they also realized that people analytics needs to be done in context for the extracted people traits to be actionable and useful. For example, a person's Openness, one of the Big 5 personality traits, may be very high in her professional life but quite low in her dating or romantic life. Thus, understanding a person's generic traits as they have done at IBM has limited value.
In addition, they aim at empowering both individuals and organizations with the same people analytics technologies to benefit both sides. In particular, such technologies will enable individuals to easily discover opportunities that match and value their individuality, while enabling businesses to easily reach their desired audience without blindly bombarding many with irrelevant information. To explore these uncharted frontiers, hopefully with more freedom, the duo decided to venture out on their own to pursue their goals:
Michelle is an accomplished computer scientist and an expert in the field of Intelligent User Interaction (IUI), which intersects Artificial Intelligence (AI) and Human-Computer Interaction (HCI) to enable better man-machine symbiosis. She received a Ph.D. in Computer Science from Columbia University. Prior to starting Juji, Inc., Michelle had led and managed the research and development of many cutting-edge IUI technologies and solutions at IBM, including context-sensitive, multimodal human-computer conversational systems, personality analytics, and opportunistic crowdsourcing. As an intrapreneur, Michelle's work has been incorporated into or created over a dozen IBM products and solutions, and resulted in over 80 scientific publications and 35 issued/filed patents. Her work has also been featured in press, including Venture Beat, Washington Post, and MIT Tech Review.
Huahai is one of the rare breed of technologists who have the expertise of a trained psychologist and a pragmatic software craftsman. He received a Ph.D. from the University of Michigan in a field that intersects Psychology and Computer Science. Prior to co-founding Juji, Inc., Huahai was a Professor of Information Science at SUNY Albany and then a Research Scientist at IBM Research and the IBM Watson Group. He has written award winning software and built psychometric and data analytics systems that have been widely used in the financial and defense sectors. Huahai also created the world's first computational model for automatically inferring human needs from text, now one of the three human trait models supported by the IBM Watson Personality Insights API. His work also resulted in over 30 scientific publications and a dozen filed patents.