General Schema of User Adaptive Systems

5 minute read

Updated: October 12, 2020

## Schema of user-Adaptive Systems

Information collection - When we start to design UAS, we need to think what info system will have about the user (available info about the user)
User model acquisition - Using this information, how can u process that information (metrics, and needa analyse them)
1. First key – relevance of the information with the situation
2. Second key – representation of the data – is it enough information to determine the world? And we need to realize that we are not seeing the whole world
3. Third – reliability – is this really capturing what the user is saying
4. E.g.
  - Is the feedback relevant to the patient and was it enough to represent the situation and the real world. Doctor talks for 30 mins but only writes one sentence. Not good representation
User model - After processing of the data you come up with a user model
- Could be db, XML, knowledge model
- Is important to represent characteristics you can use for ur application
- It ties it all together and give meaning and structure to the data we collected
User model application – thinking of ways to interfere with the user, given the user model
- Can use clustering, classification, simple analysis etc
Adapting to user - Finally user sees the interface
- Could be that it is larger, colour is different, you get more description about an item

## Main definitions

User model – data structure that contains explicit assumptions on all aspects of User that are relevant to the adaptive behaviour of the system
User model acquisition – procedure that incrementally constructs the user model and its functions to:
- Store, update and delete entries in the user model;
- Maintain user model consistency - How we update the user model etc
- Need to ensure the user model validity – it is deriving from the model hence it needs to be valid
User model application – uses the user model to:
- Make predictions and based on that you make decisions about User

Discussion – what is the most important?

Information about the user – gotta have the right information to start with, as everything is dependent on it
User model acquisition – bc you gotta collect and pick the right data
Adapting to user (interface)– when working with end users, this is what users see– HCI
User model application –
User model – user model gives meaning to the information by giving it a structure
By choosing what’s the most important they compute differently – HCI or data processing or etc
Need to have confidence in the user model

Examples – Jameson’s paper

Examples

Adapting the interface - Smart menus

We need to think about what info we have about user, diff views of the menu as user goes thru them
General schema of smart menus
1. Collects info about menus that user clicks – this is the input of the algorithm (or how long user stays, what files etc)
2. Model aquisition – register the menus that user select
3. User model – need some way to process and come up w identification of if it is a frequently used actions
  - Frequency, recency
4. Reason about user’s option choices ?
5. Modified menu content – showing the smart options

Supporting learning - SQL tutor

Learning system - We have a digital info to read, a quiz, analyse then you decide if you are progressing
- How confident they are with diff SQL commands – SELECT, ORDER BY
General schema of SQL tutor
1. User writes SQL queries
2. Application of constraints on Users solution – then it ticks how many constraints have been ticked. They write rules for each problem (constraints) to score it
3. Subset of constraints mastered/ missed by user - if the user is not good at SELECT, pull out more related problems. It decides what problem subset is appropriate for the user
4. Rule-based engine consulting the user model – how to generate the feedback and how to make a coherent piece
  - How constantly did you give a feedback?
  - And as they become better u’ll fade and give less feedback
5. Complexity and feedback – gives user feedback

Example - Amazon

Applying the schema to Amazon

Adaptative features of amazon

Amazon – today’s recommendation

What info do we have and what can we get values of
- User purchases, statements about ownerships, and ratings
  - Rating - Very small amt of users are rating. even if they rate they might rate it emotionally
  - Problem with purchases
    - Could’ve bought it as a present
    - Regular or occasional shopper – is it reliable if they only shop occasionally?
Processing algorithm
- We look at user attention, and select what’s important and converting those clicks into interest → extract into database format
User model → additional information about all the other users, bc you look at what other people buy
- Matrix of user by item – simple user model
User model application
- You choose item user is interested in from the matrix, and recommend user the items
  - And outer knowledge – other people’s interest
- By categories of the item to fetch the most common category
  - Metadata + external knowledge that tells you about the categories
- In both cases you need outer knowledge
Recommendation
- Cold start – if user hasn’t brought anything and you recommend trending products

Amazon – recommendations for all products

Same info coming about the user
Now they wanna surprise the user → in order to do that you need external knowledge about the user