While a robot may never be capable of feeling emotion, artist Erica Scourti’s bot “Empathy Deck” has the potential to pretend well enough to offer comforting help.

The premise is simple - the bot is constantly on the lookout for any tweets it deems to require an empathetic reaction. 

Following the selection of one post, the robot pulls snippets of text from the artist’s diary and a variety of sources found using the broad keyword “self-help” - creating a custom-made card designed to empathise with their situation.

So why was it made? 

Well, it was commissioned by the Wellcome Collection for its mental health exhibition - Bedlam: The Asylum and Beyond. The bot was created to explore whether therapy and caring could be automated in the future.

Does it work?

Just like any other bot, its a yes/no kind of mixed answer to that question.

Empathy Deck is chucking out a mish-mash of Scourti’s personal experiences and advice, combined with generic motivational memes pasted on wisdom cards.

The designs vary from tarot cards to the memes you see all over Pinterest. All-in-all, it’s another example of artificial intelligence not quite getting the emotional gravitas of each situation its in. But that doesn’t mean we should take away from this achievement.

Much like the company behind Apple’s Siri working on Customer Service bot that recognise when you’re getting angrier, and MIT researchers developing an algorithm to spot depression based solely on the Instagram filters that people use, robots are slowly but surely getting there.

But wait! Will this pick up the worst of humanity, like Microsoft’s chatbot?

Don’t worry. The robot will learn, but it’s not going to be left to its own devices. Empathy Deck will not end up tweeting about white supremacy because it has a lot of restrictions built in.

The bot will only interact with the people who it follows and is followed back by, and filters out any spam or offensively worded posts.

In such a situation as offering empathy to people feeling down, Scourti has also been careful to make sure the bot doesn’t respond to accounts that express what they call “genuine upset.”

While it is great work in this area, the last thing they want to do is make the account upset people further - it is just a robot pretending to empathise after all.