Antreem with Politecnico di Milano for the Wind Tre Chatbot | Antreem
Go to blog

Antreem with Politecnico di Milano for the Wind Tre Chatbot

of reading
Redazione Antreem
thumbnail

Chatbots are machines that can answer questions asked by a human in a natural language.

Although this is not a new subject in IT, over the past year the use of this type of solution to provide answers for users has been developed greatly, together with the increase in natural language conversation channels. We have seen Facebook Messenger, Whatsapp, Telegram, Instagram and others, follow a release strategy for their messaging functions on independent apps to provide a higher level user experience both under the entertainment profile, providing fun Augmented Reality filters, and to support the actual interaction with the Chatbots better.

From a technological point of view, there are numerous platforms for conceiving Chatbots without writing any code. Once the configuration and the answers have been created, Chatbots of this kind can also be activated in parallel on various channels such as those mentioned above, to which the constantly valid web channel can be added. Therefore, it is not a case of simply applying one technology, which may already be ready for the simplest cases or reaching more advanced mixed or totally customized Artificial Intelligence solutions. Rather, it is about planning the service design, i.e. finding the simplest way to satisfy users’ needs.

Antreem guided Vincenzo Bisceglia – a student from the Master’s Degree in Communication Design at the Politecnico di Milano – in designing the Wind Chatbot for his degree dissertation, in association with the company’s digital team where he was doing his work placement.

Classification of Chatbots

The study was based on a useful classification by task and complexity:

  • Editorial: through a key word matrix, they identify a response within a repository and typically send the user to consult a help section of the website
  • Customized: they allow information to be received by configuring the periodicity and a series of interests, e.g. the weather
  • Self-Care: the user accesses their account and performs actions such as changing their subscription plan, purchases and updating their personal details. The problem of cross-platform authentication arises here, i.e. connecting the conversation channel of the account with the native one of the service.
  • Mixed: through the proposal of customized information, the user receives offers of products or services. The case of searching for a flight, based on alerts, which is concluded with the purchase of the ticket.

Personalization of the conversation

A key part of the design of the conversation is personalization. This guiding principle is at the heart of the operation of Google and Facebook algorithms the aim of which is to “give the user what they want”. For Chatbots this objective is extended so that the user is given what they want in the way they prefer. The modalities can comprise both the choice of conversation channel and linguistic aspects such as the use of local dialect expressions, reaching the use of tones of voice that express emotions when the interface is vocal.

The subject has been tackled by scientists working on robots over the past ten years. To interact more easily, the machines are made to look like humans therefore the robots become androids. So, to what extend must a Chatbot seem human?

Let’s try to identify the different levels of customization:

  • No customization: this type of solution may be suitable in particular when the project wishes to impose a new type of conversational behaviour, considered more efficient.
  • Account: simple placeholders are included within the conversation patterns, which use the user’s data, such as a classic “Hi John”.
  • High level account: there are multiple contents of the conversation patterns and they are selected based on information present in the account. For example, the Chatbot changes language, even reaching the use of a dialect in full or in part.
  • Meta account: there are multiple conversation patterns which change based on information present in the account. For example, the user’s level of experience, which can be ascertained from the account metadata, can lead to short cuts making it quicker to perform a task.
  • Self-learning: using AI and machine learning, the Chatbot can decide on the level of customization based on the conversation itself and not on data contained in the account. For example, the use of an expression in local dialect can change the language or abbreviate a pattern using a shortcut.
  • Meta learning: the Chatbot learns from the user and therefore it not only adapts, but also finds new solutions to the problems posed.

The customization also takes into consideration the choice of channel, in the technological sense, as the means preferred by the user and the method of conversation: written text or voice-user interface. This is developing hugely, guided by mass products such as Amazon Alexa, Google Assistant, Cortana and Siri, which do not only relate to production faithfulness and speech comprehension, but also extend to emotional aspects such as tone and therefore fall under the area of what the user likes.

From customization to entertainment for protecting attention and trust

A consequence of customization is entertainment in that when the user converses and finds what they are looking for, the active proposal of contents or tasks to perform becomes fun in itself. We are in a highly ambitious field, but fundamental for creating habitual use, fidelity and greater user satisfaction.

In conclusion, Chatbots are fertile ground in which the challenge is to combine a User Experience vision guided by design with emerging conversational technologies.

We would like to thank the graduand Vincenzo Bisceglia, the Politecnico di Milano and Wind Tre for giving us the chance to try out the different Chatbot related projects being developed at Antreem in this specific area.

Redazione Antreem
Written by