So what is a chat-bot, and can they be a feminist? Most of us only have experience of chat-bots providing help in our online banking, drawing on an archive of possible answers trained by AI to match up with recognised questions. Cecile and Sharon explain: ‘Our feminist approach to computation means that we are not just coding for coding’s sake, we are interested in critiquing both the means of production (e.g. the process of making) and the outcomes (e.g. the software/object created). This also means that the data and the context of our coding work are important. Therefore, we are delighted to announce that we have teamed up with the Reanimating Data Project to create a chat-bot that speaks to the Women’s Risk and Aids Project archive (WRAP).’
So Friday was the first in the series of workshops and Suze worked patiently with a diverse group of participants whose expertise ranged from ‘complete novice’ to ‘rusty’. We learned about the difference between the ‘front’ and ‘back-end’ of applications and gained a sense of the painstaking work that goes into to creating the kinds of interfaces and functionalities that we take for granted. As part of setting up the task I brought along some of the original pamphlets from the study. However it was when we connected our coding to dummy data from the archive that we got a taste of what it might mean to create an interface. The chat bot asked us who we wanted to talk with in the archive. We said ‘Mary’. And then Mary appeared, ready to answer our questions: a voice from the past meeting us in the present. At this point there was an avalanche of questions about the original study, how old were they now, were they even alive? What had the original consents entailed and how it might be possible to translate them into a new media landscape. The conversation moved into our own teenage experiences, questions about what had changed in intimate relations over 30 years and how much we had in common as women from different backgrounds and cultures. We also began to understand the power of the chat-bot as a tool for a new kind of data analysis: what kinds of questions would she be asked, how would these shape her approach to the archived material. Even though we knew “Mary” was a sequence of Python statements there was still an overwhelming emotional response to “Mary” speaking back – especially for those, like myself and Sharon, who have been working on the data for the last year.
Our plan is to work towards building a chat-bot over the next two months and for this to be part of our presentation of the fruits of the Reanimating Data Project in Manchester on March 7th. We don’t know what this will look like, we expect that Suze will need to finish it off for us. We are also unsure whether the chat-bot-and-archive will need supervision – probably. We know that there are difficult stories in this archive, including stories of bad and non-consensual sex, or loneliness and self-doubt. These are also part of the historical record and while we have made sure that the interviews are anonymous we are unwilling to censor or erase the substantive content. So we just need to think carefully about the situations in which the chat-bot is used and to understand a bit more about the black box of AI that we evoke when we say the bot is trained with the data. In what ways might we need to intervene in order to encourage her to be a feminist-bot?
Yet again, the project of reanimation has brought up vital contemporary questions. Starting with an archive is an exciting prospect, opening up many opportunities for discovery, collaboration and play. I am very grateful to Sharon and Cecile for agreeing to work with the WRAP project. Not only are we developing some of the skills and awareness necessary for understanding how to make things in a digital age, we are also bringing together tools and materials in new ways that open up the possibilities of what a (feminist) chat-bot might be. I am especially intrigued to fund out what people want to ask the past as well as to discover if these questions reveal aspects of the archive that have as yet been occluded.