The Future of Customer Service: On Chatbots and Data Privacy
The world has seen tremendous developments on artificial intelligence (AI).. From simple computing tasks, AI and Machine Learning (ML) has spawned several business applications such as cybersecurity defense, human resources, market prediction, accelerated reading, accounting, Fintech, and healthcare. AI-powered chatbots are ubiquitous in the tech scene and serve a wide-array of purposes.
In the Philippines, startup Chatbots PH was recently launched to aid local businesses in responding to customer needs. Anti-sexual harassment chatbot Gabbie was also introduced by Women’s rights group Gabriela, helping women to report any incident of sexual harassment. To a large extent, e-commerce has devoted investment and research in improving the accuracy of prediction and script responses of its chatbot customer service which are commonly known as virtual assistants. To global e-commerce brands such as eBay and Sephora, chatbots have become indispensable in creatively connecting with their online buyers, increasing marketability, and boosting profits. 
Nature of AI and Chatbots
How does a chatbot work and what makes it appealing? In a nutshell, a chatbot is an AI software which can simulate a conversation, this could either be text-to-text or text-to-speech conversation, with a user in a natural language. This can be made using various media such as messaging apps, websites, or telephone. It is a cutting-edge expression of human-machine interaction limited to a question-and-answer format based on natural language processing. Through a conversational interface, AI-powered chatbots digest complex requests and customize replies – the two primary tasks of a chatbot.
ML applications are built on a neutral network which uses code for its learning algorithm. This algorithm teaches the program to use data sets such as the user’s salary or loan history and process data inputs feeded to it in order to come up with a decision. Unlike the usual code, a learning process produces results built from fundamental components of neurons and weights thereby putting emphasis on the amount and quality of information previously provided to the ML.
The success of a chatbot program depends on its ability to accurately process user request. In order to provide the right response, the chatbot must understand the content user’s request by identifying the user’s intent and extracting data and other relevant information. Expert Systems note that a chatbot response can be a generic reply, a text retrieved from a knowledge base with an entire archive of answers, a contextualized response based on data the user gave, data stored in enterprise systems, a result of an action in an interaction with backend application/s, or another question which can help the chatbot understand the request. The benefits of AI and chatbots include quick real-time response, cheaper maintenance, ease of transactions, lower operational costs, minimized errors, higher consumer engagement, better customer interactions, and versatility and capability to handle simple transactions. Through time, chatbots will be able to improve interactions. But since chatbots follow a set of rules or codes, the process takes a while before it can minimize errors and address frustrations which prompt users to press the button for a chatbot-to-human transfer.\
Legal Conversations on Chatbots
What are the legal implications of relying on chabots? With businesses recognizing the unique value of automated customer service, governments must ensure the protection of data privacy. In processing huge volumes of data which include personal data, using chatbots can trigger some legal issues such as data protection. With the GDPR in place, users or the data subjects now have rights such as the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or significantly affects him.
Scholars have also pointed out a glaring issue when it comes to AI or ML-powered chatbots in contrast with the traditional code – its decision-making process is technically a black box. With such an opaque thinking mechanism, diagnoses of patients and approval of credit applications are vulnerable to discriminatory data sets and data inputs. Weighing outcomes and filtering data inputs which are able to avoid bias decisions are gaps that have yet to be fully resolved thereby putting in doubt the fairness, reliability and ‘humanity’ of chatbots.
Potential Risks on Privacy and Data Security
Some of the hesitations in relying too much on chatbots are brought about by the possibility of compromising data and selling of personal information obtained from a simple chatbot query. From an ethical perspective, the dilemma lies in how stakeholders should tolerate possible ethical mistakes that chatbots may commit.
Programmers and retailers should be adept in the constitutional rights of the users and the possible liabilities that may arise for failure to comply with data privacy rules. With the implementation of GDPR and strengthening of data privacy laws in the country, the legal landscape on data protection urges companies and users of AI and ML programs to address the lack of transparency and the questions on autonomy by providing a detailed information on how their AI and ML-powered programs operate and process data in order to protect the welfare of consumers and avoid hefty fines.
With chatbots taking over basic and complex tasks, adaptability and accuracy of chatbots or learning machines can define the future of customer service. In the end, stakeholders and policymakers should be able to identify and address ethical and legal issues and craft mechanisms to address them in order to ensure the safety of using chatbots.
 Jim Shook; Robyn Smith; Alex Antonio, Transparency and Fairness in Machine Learning Applications, 4 Tex. A&M J. Prop. L. 443 (2018
 Ariela Tubert, Ethical Machines, 41 Seattle U. L. Rev. 1163 (2018)