CHATBOTS UNDER-HYPE: THE NEW FRONTIER TO SIMPLIFY COMPLEX KNOWLEDGE PART 4

Just as a reminder, Complex Knowledge is any combination of regulatory, statutory, legal, tax, tariff, policy and procedure matter, which is primarily found within documents.

USE CASE: HOUSEHOLDER PLANNING PERMISSION (CONTINUED)

Creating the Knowledge Map and Chatbot

The Knowledge Map is a symbolic algorithm based upon Choices, Pathways and Outcomes. By its very nature, the edge of a Knowledge Map, which establishes its boundaries, are defined by each Pathway’s end-points. An end-point is either an Outcome or a Link to another Knowledge Map.

Each Symbol within a Knowledge Map defines the type of Dialogue Operation such as Choice, Pathway, Outcome or Note. Within each Symbol there are Properties that contain the Dialogue Script. This Script maybe complemented by an in-line Dialogue Picture of Video.

As previously explained, a Picture can be used to improve the precision of a Decision.

The Video can be used for Just-in-Time training in the context of a given Dialogue-Step. This poses an interesting question of when is an expert not an expert. The answer is when they come across a Dialogue-Step whereby they do not have familiarity of the contextual knowledge. The in-dialogue video is only triggered by the person and therefore is not used when there is familiarity.

The Knowledge Map has some embedded rules such as all the Symbols need to be connected and that all of the Pathways end with an Outcome or Link.

Once the Knowledge Map is completed the Chatbot can be created through various techniques such as writing software code or through an automation process.

In terms of the Loft Extension the following Knowledge Map profile emerged:

• 42 Symbols
• 16 Choice Symbols with
• 36 Decision Options
• Planning Permission Required = 9 Outcomes, each with a different reason  
• Planning Not Permitted = 4 Outcomes, each with a different reason
• Planning Permitted = 3 Outcomes, each with a different reason
• Converted Buildings = 1 Outcome (outside scope)
• Flats / Maisonettes = 1 Outcome (outside scope)
• Other Buildings = 1 Outcome (outside scope)

It is worth comparing this type of profile data to the ‘original source’ where such data is non-existent as the Complex Knowledge is in the form of ‘dense’ documentation. 

Testing the Chatbot

Once the Chatbot is completed, the testing can commence. It is important to ensure the logic and narrative accurately reflects the ‘original source’ of the Complex Knowledge. But, let’s go back to the status quo. The ‘original source’ of the Complex Knowledge is a document. In this case, it consisted of two national documents: the regulation and the guidelines for understanding the regulations. None of these documents were subjected to usability tests. As this is the de facto standard, it is no surprise the Complex Knowledge in documented form is:

               “difficult to use and easy to misuse”

In terms of testing, one important consideration is the clarity of the dialogue during the User Decision Journey. A universal benchmark is that the dialogue can be read, understood and moved on within eight seconds. Experience has shown that the number of characters, including spaces, within a Dialogue-Step should be no more than 300. Enforcement of this limit enables the Complex Knowledge to be deconstructed and reconstructed into small simple steps, no matter the complexity of the subject.

Masking Complexity

The User Decision Journey is through the interaction with the Chatbot using a Conversational User Interface, which can be voice or text. The user does not see the Knowledge Map as the interaction is in context to the User Decision Journey. In terms of outliers, the shortest and the longest User Decision Journey can be measured in terms of the universal benchmark of 8 seconds per Dialogue-Step, which is the typical Attention Span. Thus, the following benchmarks could be established in context to the Loft Extension advice:

 Shortest User Decision Journey

• Number of Dialogue-Steps: 3
• Duration: 24 Seconds
• Outcome: Planning Permission Required

Longest User Decision Journey

• Number of Dialogue-Steps: 19
• Duration: 2.5 minutes
• Outcome: Planning Permitted

The implications of masking complexity are quite profound. The ability to mask complexity and include JIT training provides the basis for up-skilling and down-skilling, whilst spreading the ability for more and more people to self-manage inside and outside the organisation.

The Competencies for creating a Chatbot Knowledge Map

To be continued .. part 5