NextGen Intelligent Influences with Behavioral Sciences

RESEARCH AND INNOVATION

NextGen Intelligent Influences with Behavioral Sciences

 
June 15, 2020

As AI becomes central to technological systems across industries, personalized contextual recommendations and the ability to anticipate customer needs must improve. However, over time, AI systems can demonstrate a biased behavior instigated through data sets. Behavioral sciences can help understand and mitigate such biases and nudge the AI system towards a fairer analysis and more positive action. This paper throws light on intelligent nudging and how it can help track, predict and influence human behavior for efficient and effective decision making. 

Intelligent nudging: The key for positive reinforcement

Intelligent nudging enables enhanced decision making through positive reinforcement, minimal constraints, and indirect suggestions. Based on the theory of behavioral economics and choice architecture, intelligent nudging draws on people’s existing intentions, making it easier for businesses to offer better choices for enhanced customer experience. Underpinned by a Machine First™ Delivery Model (MFDMTM) and a behavioral analytics’ engine, intelligent nudging plugs in a ‘choice architecture’ with ‘data architecture’ to help create an AI-enabled recommender system. This helps offer personalized, automated, real-time recommendations to consumers based on consumer actions or transactions, consumer ‘biases’, and recommends corresponding nudges transforming the ‘market of one’ into ‘nudges for one’. For instance, intelligent nudging can help build a recommender engine that combines consumer activities, digital phenotypes and biases to help consumers understand nudged intervention or recommendation.

Companies can also leverage big data and IoT to further understand consumer preference in real-time. For instance, e-commerce companies can leverage pre-configured behavioral modeling to map consumer’s phenotypes for certain biases and recommend nudges integrated with standard AI/Deep Learning (DL) based recommendation algorithms.

Climate activist companies can also leverage nudges for recommending renewable energies, solar solutions, plastic-free accessories, greener materials and eco-friendly products. In addition, insurance companies can mine biases from consumer usage patterns to unbundle, redefine and re-bundle the portfolio for higher productivity and optimization. Financial companies can ingest digital phenotypes to identify consumer biases for offering customized financial solutions and airport authorities can design optimal gate change notifications based on passengers’ biases. Retail owners can also position micro fulfilment retail centers at optimal locations based on neighborhood behavioral biases. However, nudges have to be transparent and easy to opt out of for positively transforming customer behavior.

Leveraging behavioral modeling to mitigate bias in AI

This requires implementing AI systems where datasets are trained for no bias. Supervised AI can lead to inaccurate or biased predictions and unsupervised AI can produce false patterns. For instance, in generative AI models, inaccurate sampling can amplify noise in its probability distribution and in cognitive AI or neural nets, bias can penetrate through the network bias. This impacts the AI system’s decision-making capability. Behavioral modeling of the training data helps incorporate good bias, assess and categorize bias, identify the ‘bias’ with better algorithmic outcomes and determine bias variance in the sampled population principle. AI systems underpinned by behavioral science can support fair and positive action, enabling better handling of inherent system bias in machine learning or deep learning applications.

Paving the way for responsible decision making with new nudges

Nudges that harness the collective power of creativity, technology, and data can result in the next generation of intelligent influence. Powered by AI, the “new nudges” can significantly enhance recommendation engines, and lead to more responsible and trustworthy decisions in artificial, autonomous and intelligent agents, enabling better business outcomes.

Shampa is currently pursuing research in the area of Quantum Computing, and is working with the TCS Research & Innovation (R&I) team towards building the Tata Quantum Computer interface. She holds a Ph.D. in Experimental Physics.

https://in.linkedin.com/in/shampa-sarkar-48630023