Virtual Assistants and Ethical Implications

*Abhishek Kaul*

## **Abstract**

Virtual assistants are becoming a part of our daily life, from our homes to our work. Sometime we may not even know that the customer service agent that we were speaking with is a virtual assistant. These assistants continuously collect information from our interactions and learn many things about us. The information they gather over time is enormous. This chapter introduces the concept of ethics, discusses the ethical principles of virtual assistants, (Transparency, Justice & fairness, Nonmaleficence, Responsibly and Privacy). Although there is limited regulation governing these virtual assistants, practical guidelines and recommendations are provided for designers and developers to understand the ethical implications when building a virtual assistant. In this chapter, we also discuss technology and learning techniques for virtual assistants and present examples on how to ensure ethical virtual assistant.

**Keywords:** virtual assistants, artificial intelligence, ethics, deep learning algorithms, natural language processing, natural language understanding, fair AI, transparent AI

### **1. Introduction**

Organizations are rapidly deploying Virtual Assistants aka bot technology [1] for automating communication, customer service, conversational commerce, product recommendation, education support, financial services, medical services, entertainment, social outreach and self-service tasks. They offer 24/7 service and fulfill the need of millennials [2] for real time responses. Virtual assistants enable organizations to reduce costs, increase brand loyalty and better serve customers. However, virtual assistants are built by humans using artificial intelligence (AI) technologies and have wide ranging ethical implications which are important for organizations and consumers to understand.

Many countries have published AI policy guidelines [3, 4]. These guidelines provide a broad level objective for the use of AI – to ensure human-centric, safe, and trustworthy AI. One of the most important aspect in all guidelines is ethics, "AI should be ethical, ensuring adherence to ethical principles and values". Although, AI by its very nature is a form of statistical discrimination (finding patters in data), the discrimination becomes objectional when it places certain privileged groups at a systematic advantage and certain unprivileged groups at a systematic disadvantage. For example, the loan application algorithm gives higher credit scores to older males due to training bias. Objectional discriminations can arise due to multiple reasons like wrongly defining the business objective [5] of machine learning model, using unrepresentative data or data with existing prejudice [6] for training or by selecting wrong attributes or features of the AI model.

#### *Virtual Assistant*

Significant work has been done in the area of Ethically aligned design for Autonomous and Intelligent systems by IEEE [7]; and in the area of Facial recognition technologies [8]; but the area of virtual assistants has seen limited guidelines or regulation. California "Bot Bill [9]" provides only limited protection for consumers in terms of bot self-declaration.

In subsequent sections of this chapter, we first define "What is ethics?" and then discuss on ethical principles for virtual assistants. These principles provide key ethical considerations that designers and consumers should understand. Next we discuss different types of virtual assistants deployed today deep, dive into the technology and learning techniques that make them ethical. Further, we analyze the guidelines and legislations that companies and governments have published. In the last section, we look into what is the probable future of super intelligent virtual assistants.
