Voice technology operates on Natural Language Processing algorithms. Studying these algorithms is often a pervasive task, as they exist ‘out there’ in the cloud. However, they can be researched as algorithms often echo the social structure, software structure and physical structure in which their creators are embedded.
N
Author
Patrick Hondius
NLPs are often provided by large corporations that aim to commodify data. The commodification of data can be negated through creating a neutral NLP that is not related to any of the cloud computing giants.This research was conducted under my supervision on behalf of Mirabeau, a Cognizant Digital Business. With a team of young talented master and PhD students in the human sciences, we have conducted foundational research into consumer behavior in the focus industries of Finance and Travel, and into the role that new technologies play in people's daily lives. Great research is the precondition for great design.
The current GDPR regulations allow for the storing of sensitive data (European Commission for Justice and Consumers, 2018: 5). By encrypting data, appointing data protection officers and making data protection impact assessments, the commission for justice and consumers deems companies qualified to handle sensitive data (ibid.: 6-7). These factors should be taken into account, when creating a neutral NLP. In this way, the data is no longer the object of revenue, but rather the voice interaction and getting to know the user.
By personally knowing the user, the company providing the voice assistance can tailor their assistant to fit the user’s needs and expectations. Users should also be notified by what data is collected.
The Dutch DDMA (Data Driven Marketing Association) have taken the initiative to create a hallmark for safe and good voice assistants. Although it is still in the making, this venue would provide companies and users with clear information about the dataflow behind a specific voice assistant. Users will benefit mostly from this, as it could become a medium that clarifies terms of services and other legal texts that are difficult to understand for most.
The algorithm exists solely to analyze text. The steps in a conversation are made by the developer. The hard thing about that is remembering the exact context of the conversation and knowing what the user expects.
The interaction with a voice assistant has an element of sensemaking to it. If the primary interaction of users with an assistant leads to a negative experience, users do not return to the assistant. The sensemaking of users should be understood as them bringing their social and cultural background to an interaction. Similar to each person having different needs and expectations, the users can have different social and cultural backgrounds.
Therefore, to maximize the support that a voice application gathers, testing should be done with a socially and culturally diverse group of testers. When testing is done accordingly, users will be able to voice their concerns and (dis)content about the voice application. Users should be prompted by a voice application with the commands they can use to navigate the application, as there is no directory of available commands. In the test-phase these prompts could provide a good starting point. Lastly, creating a voice assistant is a prolonged process of making continuous adjustments.
The dynamics and versatility of the Dutch language, can cause tiny discrepancies in the workings of a voice application. This can be adjusted by the developers. Syntax and semantics should therefore be examined closely, when adjusting the intent recognition of a voice application. These adjustments can only be made when an assistant is ‘live’.
Therefore, voice applications should be viewed as a long-term investment, as they require continuous and prolonged maintenance. By researching, the user’s perspective on and interactions with VA to ‘‘…to gain a deeper understanding of consumer judgement and behavior towards brands.’ (Mari, 2019: 1) This long-term investment can be turned into a profitable venture, in which data is no longer the primary commodity, but rather the service or product of a client company. On a broader scale this can be achieved, by studying the interplay between customer, brands and retailers’ behaviors in response to “machine behaviors” (Rahwan et. al., 2019). As in that manner the user’s understanding and behavior can be researched.
This article is an abstract of the research report. Interested in reading the full report? Please contact me, info@henkhaaima.com.
Read other articles
For a long time, chatbots were mainly notorious for what went wrong. Quietly, a lot has been improved since then. Patents point to an imminent leap forward. With better trained AI, real-time interaction and more personalization, conversational AI can be used in more and more ways.
Author: Patrick Hondius
N
NLPs are often provided by large corporations that aim to commodify data. The commodification of data can be negated through creating a neutral NLP that is not related to any of the cloud computing giants.
The current GDPR regulations allow for the storing of sensitive data (European Commission for Justice and Consumers, 2018: 5). By encrypting data, appointing data protection officers and making data protection impact assessments, the commission for justice and consumers deems companies qualified to handle sensitive data (ibid.: 6-7). These factors should be taken into account, when creating a neutral NLP. In this way, the data is no longer the object of revenue, but rather the voice interaction and getting to know the user.
By personally knowing the user, the company providing the voice assistance can tailor their assistant to fit the user’s needs and expectations. Users should also be notified by what data is collected.
The Dutch DDMA (Data Driven Marketing Association) have taken the initiative to create a hallmark for safe and good voice assistants. Although it is still in the making, this venue would provide companies and users with clear information about the dataflow behind a specific voice assistant. Users will benefit mostly from this, as it could become a medium that clarifies terms of services and other legal texts that are difficult to understand for most.
The algorithm exists solely to analyze text. The steps in a conversation are made by the developer. The hard thing about that is remembering the exact context of the conversation and knowing what the user expects.
The interaction with a voice assistant has an element of sensemaking to it. If the primary interaction of users with an assistant leads to a negative experience, users do not return to the assistant. The sensemaking of users should be understood as them bringing their social and cultural background to an interaction. Similar to each person having different needs and expectations, the users can have different social and cultural backgrounds.
Therefore, to maximize the support that a voice application gathers, testing should be done with a socially and culturally diverse group of testers. When testing is done accordingly, users will be able to voice their concerns and (dis)content about the voice application. Users should be prompted by a voice application with the commands they can use to navigate the application, as there is no directory of available commands. In the test-phase these prompts could provide a good starting point. Lastly, creating a voice assistant is a prolonged process of making continuous adjustments.
The dynamics and versatility of the Dutch language, can cause tiny discrepancies in the workings of a voice application. This can be adjusted by the developers. Syntax and semantics should therefore be examined closely, when adjusting the intent recognition of a voice application. These adjustments can only be made when an assistant is ‘live’.
Therefore, voice applications should be viewed as a long-term investment, as they require continuous and prolonged maintenance. By researching, the user’s perspective on and interactions with VA to ‘‘…to gain a deeper understanding of consumer judgement and behavior towards brands.’ (Mari, 2019: 1) This long-term investment can be turned into a profitable venture, in which data is no longer the primary commodity, but rather the service or product of a client company. On a broader scale this can be achieved, by studying the interplay between customer, brands and retailers’ behaviors in response to “machine behaviors” (Rahwan et. al., 2019). As in that manner the user’s understanding and behavior can be researched.
This article is an abstract of the research report. Interested in reading the full report? Please contact me, info@henkhaaima.com.
Read other articles
Read other articles
