Some Trustworthy AIs at MLH Local Hack Day
Solutions proposed at the hackathon
There are multiple ways technology can prove to be threatening, especially in a world where privacy of individuals is being challenged and played with by various businesses and/or for technological experiments. Understandably so, AI can seem suspicious and untrustworthy.
Srishti Institute ACM SIGCHI Student Chapter and Interaction Design Foundation (IDF) community had organized MLH’s (Major League hacking) Local Hack Day (in collaboration with Github and Microsoft) on 1st Dec 2018 at Srishti Institute of Art, Design, and Technology, Bangalore. The theme of the hackathon was “Designing Trustworthy AI”. Participants through their interventions explored ways to understand and challenge how trust could be imbibed as a design principle while designing an AI product.
Here are some of the solutions that the participants came up with so that anyone who is exploring what it means to build a trustworthy AI can learn from people who plan to do the same! :
1. Counterfactuals in Machine Learning
By Avi Jain
They built a script that takes in a particular instance, a machine learning model, and provides an example closest to the given instance for which the prediction of the model changes.This can be explained from an example:
“Peter applies for a loan and gets rejected by the (machine learning powered) banking software. He wonders why his application was rejected and how he might improve his chances to get a loan. The question of “why” can be formulated as a counterfactual: What is the smallest change to the features (income, number of credit cards, age, …) that would change the prediction from rejected to approved? One possible answer could be: If Peter would earn 10,000 Euro more per year, he would get the loan. Or if Peter had fewer credit cards and hadn’t defaulted on a loan 5 years ago, he would get the loan. Peter will never know the reasons for the rejection, as the bank has no interest in transparency, but that’s another story”
2. AI for Protecting User Privacy
By Deepanshu Setia and Archana Rao
They multi-class image classifier with CNN model that allows users to protect their image data from being shared across various platforms while using an NLP powered solution for secured contacts.
3. Smart Glasses for Autistic Persons
By Micah Alex and Kirti Kaushik
Autistic children often experience sensory overload causing panic attacks and meltdowns. With a speculative design approach, the team proposes smart glasses called “Clutterfly” that targets noise and visual overload by integrating AI, VR, and noise canceling features. Various guidelines to ensure trustworthiness were also proposed such as relationships of transparency and incremental independence with the product.
4. Meanbot — A Sarcastic Chatbot
By N Chandrasekhar Ramanujan, Jahnavi Vegad,
Smart Assistants use smart voices and are evolving to deliver solutions to make our lives better. However, the AI assistants these days seems to have a machine-like, mechanized personality which is harder for humans to trust. The team designed a “MeanBot” to propose the idea of having a smart assistant with a more trustworthy personality. The personality is meant to get more personalized and evolve using NLP as per the personality of the user using it.
5. Maps and Trustworthy UX for AI
While evaluating the experience and the state of driving on Indian roads, their project aims to bring trusted experiences to navigation on unfamiliar roads in the Indian context. “Alert Mode” is a feature proposed for Map applications like Google Maps or OSM to improve driving experiences more accurate and realistic in accordance with realities that can’t be examined by satellite cameras such as potholes, unsafe areas or lack of functioning street lights due to recent power outage, etc using local data such as reports by people, newspapers or radio and TV telecasts. You can find out more about their project on this link.
6. Hife
By
They proposed an intervention- “Hife” with the aim of shifting the perspective of fitness from the end result to the journey of achieving it using AI and suitable interactions that help in an individual’s overall growth.
7. Specialized Cooking training using AI experience
By Achyuth, Nehal & Joe
They designed an AI experience to train users step by step cooking instructions and generate the recipes-based dishes based on ingredients available at home.
8. Intelligent disease prediction system
By Rohith Kumar S, Suraj Srinivasa
They created “Intelligent Disease Prediction System” based on Naive Bias Algorithm that is useful to predict the patient’s disease and state based on provided symptoms that are based on a historical symptoms dataset.
9. Shop Assist
By Meghana R Upadhyaya, Jyotirmayee Das, Ketaki M Paranjape
They made an AI application called “Shop Assist” that allows you to scan clothes while shopping to determine their quality on factors such as weaving defects, texture defects, dye-color defects, etc. This application using image processing and AI helps in augmenting the showroom shopping experience while also helping the user connect with online fashion e-commerce for relevant/ better recommendations.
10. AIDO- an AI To-Do App
By Bhushan Sharma, Chanasya Arora, Prem Sameer
“AI To Do” or AIDO is a smart AI assistant that interfaces with user to inform consent, enable user to create tasks/labels, explore suggestions and view/edit personal data. The AI interacts in the form of avatar using animation, voice and text using human like conversations. Control is given to the user for viewing decision information based on the labels, periodicity, preferences.. etc and also allows users to edit them enabling transparency of the application.
11. Minutes of Moments — a Meeting Transcribing App
By Kashvi Bajaj ,
They created an AI assistant that helps users take moments of meetings with ease. The app not only converts speech(from your meetings) to text but also structures the text and summarises key points and takeaways from the transcribed text.
12. Predictive household items ordering service
By Harinie Ananth, Saili Gupte
Quite often, we run out of things and we make a note in order to buy those things in the future. Their proposed intervention aims to augment and simplify the household care by predicting the need for buying new items and helps you order those items accordingly.
13. Les Femme
By Suhail, Aniket
They proposed an app- “Les Femme” that helps women track their periods and make available resources and facilities according to their needs.
