Allow for sudden increase in the heat pump awareness

[charlotte-avery]

Add heat pump awareness campaign intervention:

• Intervention is set up to model a sudden increase in heat pump awareness throughout the population as a result of, for example, some widespread campaign.
• User inputs the date of the campaign (heat-pump-awareness-campaign-date) and the heat pump awareness achieved due to the campaign (heat-pump-awareness-due-to-campaign).
• The heat pump awareness can never decrease from the initial heat pump awareness (heat-pump-awareness) even if the user uses a lower value for heat_pump_awareness_due_to_campaign.
• Agents in the population are randomly assigned to be either aware or not aware, before the campaign and after the campaign, with a probability given by heat-pump-awareness and heat_pump_awareness_due_to_campaign, respectively. This logic is captured in the functions: construct_population_awareness and construct_population_awareness_due_to_campaign.

Code no longer updates to 100% awareness at the boiler ban announcement date:

• Instead, 100% awareness only happens after the ban is in place.
• If this is effect is desired, one can set the campaign date to the boiler ban announcement date, and set heat_pump_awareness_due_to_campaign=1.

[shengy90]

Is this the right variable? Should you have the at_current_step suffix?

[charlotte-avery]

This is referring to the property DomesticHeatingABM.heat_pump_awareness defined in models.py which returns households_heat_pump_aware_at_current_step / household_count, which is what I want.

[shengy90]

the self.is_heat_pump_aware logic is: either (already has heat pump) or (is heat pump aware) because by definition, if they have a heat pump, they must be heat pump aware.

I'm not sure if the same logic follows into is heat pump aware after campaign:

• this variable suggests that this household is aware because of the campaign
• not 100% sure 1) if this variable name actually reflects the variable you're trying to create 2) how this variable will be used later on and whether or not it reflects the variable name

will revisit this comment later

[shengy90]

After having read the entire PR, I now understand this variable to be latent variable that gets activated later on if the campaign date is reached. But i'm still unsure if we need this variable to implement the campaign effect..

[shengy90]

Just to check:

heat_pump_awareness_due_to_campaign: the target awareness we're trying to get to with the campaign
heat_pump_awareness: the current level of awareness in the absence of the campaign, i.e. baseline awareness?

This function basically tries to select the people who are currently not yet aware, and randomly assign households s.t. after the campaign the total number of 'aware' household = specified target?

If i'm summarising correctly, please could you:

1. Make sure that the doc string correctly summarises what the function is trying to do, including the argument definitions, as it's not very clear.
2. If my understanding is correct, is heat_pump_awareness basically the proportion of population_heat_pump_awareness that is True? If so - can we clarify this in the code as it's a bit unclear.

[charlotte-avery]

Yes your understanding is correct. I will update the doc strings.

[shengy90]

This comment and implementation is unclear to me.

The comment #Awareness cannot decrease over time suggests that these variables are 'changing' over time but isn't heat_pump_awareness_due_to_campaign and input to the simulation (i.e. the target) so it can't change over time. What changes over time is the awareness _at_time_step?

I think the code will be a lot clearer if this logic is elsewhere (possibly right at the beginning of the script where simulation inputs are parsed). Options:

1. or you output a "warning" to in the log to say heat_pump_awareness_due_to_campaign value is overriden with heat_pump_awareness because value is lower or error or
2. don't let the simulation run and warn the users of invalid simulation input

I think the soft overriding here is very implicit and unclear and I'm trying to guess what you're doing.

[shengy90]

Do we need to go through all these trouble? Why couldn't we simply set heat_pump_awareness value to the heat_pump_awareness_due_to_campaign value and randomly assign households the target awareness value?

My understanding of create_household_agents function where this function construct_population_awareness_due_to_campaign is called:

• create_and_run_simulation is the handler function
• at the start of simulation, this function will create the model (the universe)
• will create households, and the households would have been assigned heat_pump_aware at the very beginning
• then model.run will collate all the collector functions, and update the households at each timestep

If you want to model heat pump awareness that will increase over time due to the campaign, then shouldn't all this logic live in collector.py, rather than right at the beginning?

[shengy90]

Although I see that in agents.py you're only setting the the attribute is_heat_pump_aware to this heat_pump_awareness_due_to_campaign after the campaign date.

Which suggests that this heat_pump_awareness_due_to_campaign is a 'latent' variable that doesn't get activated until after the campaign date.

In which case:

• I'd recommend naming the variable to something clearer e.g. will_become_heat_pump_aware due to campaign or something like that or heat_pump_campaign_target_audience or something to that effect
• And actually explain this in the doc strings because I'm just trying to infer/ guess what you're trying to do and there's no context elsewhere in the code that actually lets me confirm what I think you're trying to do!

[shengy90]

And if my guess of what you're trying to implement is correct, I feel like it would be more natural to implement this logic inside agent.py instead of creating this value right at the beginning. Creates less attributes/ methods.

All this logic could have been in update_heat_pump_awareness function - just access the heat_pump_aware_at_current_timestep value you've created, calculate the gap (target - currently aware), and flip households accordingly to achieve the target.

[charlotte-avery]

The reason I have defined awareness in model.py r.t. agents.py is to ensure that once an agent becomes HP aware, it will always be HP aware. The simulation runs by looping over timesteps for the model horizon, and then at each timestep, loop over all agents (i.e., call functions in agents.py). If we change information about agents in agents.py in within this for loop, we need to ensure that information is collected and stored somewhere so that at the next timestep, the agent that became 'HP aware' at the previous timestep, is still remembered to be 'HP aware' at the next timestep. In the current ABM implementation, this is achieved by defining whether a household is heat pump aware is defined outside this for loop in create_household_agents in model.py. Agent properties e.g., EPC, heating system, HP awareness etc. are defined in create_household_agents outside the for loop which makes sense to me.

Given that we can access the full population statistics in create_household_agents (we only see one agent at a time in agents.py), it makes sense to me to look at who is HP aware, who is not HP aware here, and then define which of the non-aware HP agents will switch during the campaign. Also, given that 'baseline' or initial HP awareness is defined here in the current ABM implementation, it seems natural to me to also define 'future awareness due to campaign' here.