PRE2019 3 Group7

Group Members

Name	Student ID
Daan Schalk	0962457
Job Willems	1003011
Jasper Dellaert	1252454
Sanne van Wijk	1018078
Wietske Blijjenberg	1025111

Subject

Use a simulation to optimize the revenue of a bar or club.

When looking at Stratumseind these days, clubs and bars that were once flourishing look empty and silent. For a manager of such a bar, it may be difficult to decide what to improve to regain the glory of those golden days. Humans are complex, so their happiness and willingness to spend may depend on different factors combined in a specific way. Because of all these different factors, experimenting with them in real life is impossible work. On top of that, change in revenue might not be seen in a week. To be sure about the effect of a factor, a longer observation period is needed. However, if you need to test a lot of different factors in a lot of different combinations, this can take ages.

That is where a simulation of the behaviours of the people in a bar or club might help. The simulation will make it easy to experiment with different factors, and observe the effect of those factors on the behaviour of the people in the bar. It also adds the possibility to run multiple simulation in sequence. This way, certainty can be reached about the influence of a certain factor on the revenue, and managers of clubs and bars can come to decisions about what to improve to make their bar great again. The average club derives around two-thirds of its revenue from the sale of beverages, and it is therefore a vital source of profits . Therefore the simulation will be focusing on how to maximize the selling of beverages.

Objectives

• Main goal: assist the owner of a bar or dance hall with ideas on how to improve customer revenue.
• Construct an AI multi-agent simulation with which we can find what variables with what capacities are needed to maximize generated income.
• Analyse the reliability of such a simulation by discussing the observations of the client and the results of previous research done in this field (if any apply).
• Analyse the possibilities and shortcomings of such a simulation.

Research variables

Note: during the early phases of this project, variables are subject to change and may still be included/excluded.

General variables

Income

• Description: The amount of money spent by patrons inside the establishment during the evening/night in order to buy alcohol. Alcohol is the only product available, which has to be bought using euros as currency. At the start of the simulation, this variable will start at a value of 0 euros.
• Used in simulation: yes
• Reasoning: Income is a crucial variable, as the problem statement is based around creating as much income for the establishment as possible. All other variables used during the simulation do relate to the generated income to some extent.

Crowded level

• Description: Is the establishment filled with patrons or are there barely any people? An upper limit of the amount of people that can comfortably fit inside the bar/dance hall has been estimated (using room size and patron experience). The crowded level is defined as the ratio between the current amount of patrons and the upper limit, displayed as a percentage. At the start of the simulation, this variable will start at a value of 0 percent.
• Used in simulation: yes
• Reasoning: As the patrons' happiness can decrease significantly when the bar is overcrowded, it is important to weigh in this factor when evaluating the latter. Patrons can also become less happy when there are barely any other people inside .

Dancing crowd

• Description: Patrons can start and stop dancing at any time inside the establishment. The Dancing crowd is the amount of patrons that are currently dancing according to their patron state. At the start of the simulation, this variable will start at a value of 0.
• Used in simulation: yes
• Reasoning: This variable does not relate heavily to other variables. It is however used when a patron has to decide whether it will start or stop dancing. Because the ability for patrons to dance is an important element of the simulation, the dancing crowd is included.

Environment specific variables

Room size

• Description: The surface size of the singular room in which all patrons can be found inside the establishment. This room can be used for all patron activities, and is also the location in which alcohol is sold. The specific geometry of the room is neglected. The surface area size is given in cubic meters.
• Used in simulation: no
• Reasoning: While the size of the room plays a role when analysing an average day at a bar, it is often not the actual size that counts, but the percentage of the establishment that is filled with patrons. For this reason, the room size is accommodated inside the crowded level variable.

Amount of bathrooms

• Description: A bar or dance hall one or multiple toilets for both male and female patrons. While this is not obligated by the Dutch law to do so (for smaller establishments), an average bar or dance hall will have multiple toilets. This variable does not discriminate between toilets catered to women or men.
• Used in simulation: no
• Reasoning: It is determined [SOURCE?] that the amount of bathroom breaks or the amount of toilets does neither directly nor indirectly change alcohol sales. When a bar or dance hall does not have any toilets, then that might cause customers to not return to that establishment on subsequent nights. Because this research only analyses a singular regular day, it does not affect the simulation results.

In this study it is shown that although people consume less alcohol during bahtroom breaks, bathroom breaks are so short that the "lost time" is easily compensated after the break, thus the total effect of bathroom breaks on alcohol consumption is zero.

Music

• Description: A bar or dance hall can play music throughout the day. It is assumed that the music genre that is played is appreciated by the patrons. The volume of the music can change, which can alter the behaviour of the patrons. The music can also be absent. This variable describes the volume of the music, which will start at a value of 0 dB.
• Used in simulation: yes
• Reasoning: Music volume can greatly change the behaviour of patrons . While the genre of the music being played could also alter this behaviour, the exact relation could not be grounded by previously done research, and is thus omitted from the simulation.

=> It could though. Studies found that when playing drinking songs in a bar, the duration of stay and spending are both increased .

Amount of active employees

• Description: The amount of people currently working in the establishment. While an employee can have different occupations within the bar/dance hall, all of them are generalized under this variable. The employees that are included are all people with occupations that require them to actively engage with the patrons. This includes alcohol vendors situated at the bar and all other employees that are able to assist any patron directly inside the establishment.
• Used in simulation: no
• Reasoning: It is assumed that the amount of active employees does not affect the patron behaviour as long as it exceeds a minimum amount, which is dependent on the amount of patrons. Above this limit, all customers can get assistance without having to wait (too long). Below this limit, patrons might become agitated as they might have to wait for long periods of time in order to get the assistance they require. It is assumed that the owner of the establishment always has enough active employees to exceed this threshold, as to not decrease the patron happiness. This variable can therefore be left out of the simulation.

Beverage stock

• Description: The amount of alcoholic beverages the establishment has in stock ready for sale. Complying with the research constraints, all types of beverages are generalised as one singular type. It is assumed that all patrons are allowed to buy these beverages and do to a certain extent enjoy consuming them. It is also assumed that all products are up to both legal and patron standards, making all of them sellable.
• Used in simulation: no
• Reasoning: It is assumed that the owner of the establishment has experience and can accurately determine the amount of beverages that are being consumed on an average day. This means that all patrons are able to buy whenever they want. For this reason, this variable does not affect the beverage income and is not included in the simulation.

Amount of seats

• Description: A bar or dance hall often has seats for patrons to sit on, but almost never enough bar-stools or chairs for everyone. When a patron is tired or does not want to stand, he or she can sit whenever a chair is still available.
• Used in simulation: no
• Reasoning: While patrons often try to sit down when they are tired, it does not replenish their energy significantly. Because the chair do not affect the energy level of the patrons, it does not affect the generated income. For this reason, this variable is not included in the simulation.

=> people do like relaxing and taking a break from the pushing and shoving, comfortable seating and individual seating are preferred above bar stools .

Food availability

• Description: Is it possible to buy food inside the establishment? Food can range from snacks to real meals, but is generalized as one undefined food item, assuming that all patrons are satisfied with this when requesting food. Food could generate more income besides beverages.
• Used in simulation: No
• Reasoning: Most bars or dance halls in the Netherlands do not sell food items. Food is ignored in the simulation in order to simulate an average bar or dance hall. Having food as a second source of income would also drastically complicate all variable relations, increasing the scope of the research. Keeping the scope of the research realistic is another reason for omitting food from the simulation.

Specified dance area

• Description: The establishment can have a specified dancing area. When such an area is available, patrons will not dance on anywhere else. The specific surface geometry or location within the establishment is ignored. The surface area size is given in cubic meters.
• Used in simulation: no
• Reasoning: There are no known differences in patron behaviour when a designated dance floor is available. It is therefore unnecessary to implement this variable in the simulation.

Patron specific variables

Money

• Description: All patrons can carry money, euros specifically. They can then only spend this money on alcoholic beverages. A patron can not spend more money than what they personally have: they can not lend money from other patrons. While real-life patrons sometimes pay for a group, this is not possible in the simulation, as it is assumed to not affect its results.[Buying drinks in rounds does affect amount of beer one consumes,] Patrons have a random amount of money on them when entering the establishment, given to them using a [TODO] distribution [mean, var?].
• Used in simulation: yes
• Reasoning: It is essential for the patrons to have money, as it is necessary in order for them to buy alcoholic beverages, thus generating income.

Willingness to pay

• Description: How easy do the patrons part with their money in exchange for alcohol? Whenever a consumer has less money to spend, he/she is less likely to spend its remaining euros. The value of this variable thus decreases whenever the patron has less money, which might cause him/her to take more time before buying more drinks.
• Used in simulation: no
• Reasoning: While this might have an effect on the amount of money that the patrons will spend throughout the day, this effect would not be large. It is also not proven whether patrons actually decrease their spendings when they have less money. They will likely be under influence when this happens which might decrease their good judgement.

Has alcohol

• Description: Describes whether the patron currently has alcohol on his/her person. A patron will not buy any more alcohol whenever he/she already carries alcohol. When carrying alcohol, the patron might consume parts of it, which slowly intoxicates him/her. Whenever the alcohol is completely consumed, the patron has no alcohol anymore. The variable is defined as a percentage, with 100 percent denoting a full beverage, and 0 percent the absent of any drink.
• Used in simulation: yes
• Reasoning: In order to generate income for the establishment owner, it is important that the patrons can buy beverages. It would be unfair if they could buy drinks and not consume it afterwards. Furthermore, better mimics reality, as real life patrons generally consume their products as well.

Intoxication

• Description: A patron can become intoxicated when he/she consumes alcohol . This variable is represented as a percentage, with 0 percent representing a completely sober individual. When the patron consumes enough alcohol and reaches 100 percent, he/she will leave the establishment immediately. Intoxication changes the behaviour of a patron .
• Used in simulation: yes
• Reasoning: Intoxication alters the behaviour of the patron significantly. It also might cause a patron to leave early whenever they become too intoxicated. For these reasons, this variable is implemented in the simulation

Alcohol tolerance

• Description: This will show how much a certain patron can tolerate drinking alcohol. This is different per type of patron, some people can better stand drinking alcohol whilst not completely going drunk .
• Used in simulation: simplified
• Reasoning: This will not be used to its fullest in the simulation, it will probably be represented by a constant value which is the same for each patron.

group sizes

• Description: This will represent with how big of a group you are there. Since it will be more enjoyable to a patron, when the patron can converse and dance with more people he knows .
• Used in simulation: no
• Reasoning: At this moment it will not be included, however when the simulation would allow it this could be included as another variable.

Happiness

• Description: Happiness will depict how much a patron is enjoying himself. This can changes because of other variables in the simulation, for example when dancing and talking happiness will increase. This can be represented as a percentage in each patron, which will have certain threshold for leaving or staying.
• Used in simulation: yes
• Reasoning: With this it can be modeled how people are enjoying themselves. Which is an important part of the simulation. Since with this can be determined if people will stay when they are happy or go when they are unhappy.

Dance affinity

• Description: Dance affinity is here to show how much a certain patron enjoys to dance. Certain people will have a higher tendency to dance, whilst others will only dance in the right situations and circumstances.
• Used in simulation: simplified
• Reasoning: This variable will be relatively small and would prove to complex to model currently. Also since the variable is likely to change in the simulation for a patron, due to for example drinking alcohol and thus becoming more loose. This will however be represented by a the same constant value in each patron.

Energy

• Description: A Patron will have energy, with which they can do actions. When they do an actions, energy is subtracted from the energy of that patron. Possible action for example are dancing, standing, talking, drinking and buying. Which all have different energy costs. This will likely be used as a certain value from which patrons can differ from, with a small difference from the mean.
• Used in simulation: yes
• Reasoning: This variable is needed in the simulation, since people who are tired are likely to go from the bar.

Constraints

Constraints are limits that are set for the simulation. Constraints are needed in order to create a working simulation. Without these constraint a simulation would prove too complex to make, therefore these constraints have been made.

Bar

• Bar cannot be remodeled or renovated
• There is only one type of alcohol/drink
• Not taking into account other bars/competitors
• Given one night with certain amount of people inside
• Bar will not sell food
• No animals are allowed
• Drinks will not expire
• The bar only has one room in which people will dance and drink

People

• People will not start fights
• People will not complain
• People will not break stuff
• People will always pay and not steal
• People will go when they are unhappy or when they have no more energy
• People do not go to bathroom

Patron states

• Drinking: The only state that can be done simultaneously with any of the other states: In this state, the patron's alchohol will be drunk, increasing its intoxication level.
• Dancing: The patron is currently dancing and socialising with others. This actions does cost more energy than any other action.
• Talking: The patron is conversing with others, This does cost some energy.
• Buying: The patron is currently buying more alcohol for him/herself.
• Standing: In this state, the patron is inactive and unsocial. This state barely costs any energy.

Users

The main stakeholder for this simulation are bar/club owners. Since the simulation takes into account which factors a bar/club owners has any influence on, it can be a great tool for giving meaningful suggestions to generate more profit. It can also be used by a bar/club owner to see if an idea he has to generate more profit has potential or not, since all configurations of factors can be tested. Since it focuses mainly on how to maximize the sale of alcohol, it could also be useful in other contexts where selling alcohol is the main way of making profit, as long as the setting itself does not differ too much from the bar setting specified in the simulation. After all, differing from the defined setting could possibly introduce factors which change the result of the simulation. The simulation could even be useful in settings like student parties, where people want to get others as drunk as possible. Question is if it is ethical to use such a tool to get people drunk.

A secondary user for which this simulator might prove useful, are the people going to the bars and clubs. Assuming that customers spend more money when they enjoy themselves and leave when they are unhappy, using the results of the simulator to improve bars would mean that bars become more enjoyable. This is beneficial for the customers: they will have a better time.

State-of-the-art

Artificial Intelligence (AI) is used more and more in (environmental) modelling. It can mimic human perception, learning and reasoning, which can be used to solve complex problems. There are a lot of difference AI techniques: case-based reasoning, rule-based systems, artificial neural networks, genetic algorithms, cellular automata, fuzzy models, multi-agent systems, swarm intelligence, rein-forcement learning, hybrid systems, and, arguably, Bayesian networks and data mining

Case-based reasoning (CBR)

This type of AI assumes similar problems have similar solutions and recalls similar past problems to solve a problem . CBR involves four steps :

• Retrieve the most relevant past cases from the database
• Use the retrieved case to produce a solution of the new problem
• Revise the proposed solution by simulation or test execution
• Retain the solution for future use after successful adaptation

Pros: by updating the database, a CBR system continually improves its performance. It can also handle large amounts of data and multiple variables and organises experience efficiently .

Cons: if there are no similar past cases, CBR cannot draw inferences about the problem. On top of that, CBR is a black-box approach and offers little insight into the system and processes involved .

Rule-based systems (RBS)

This type of AI solves problems by rules derived from expert knowledge . The rules consist of an if-part and a then-part: if (certain condition) then (do this action). These rules are fed to an inference engine, which has a working memory of information about the problem, a pattern matcher and a rule applier. The pattern matcher consults the memory to decide which rules are relevant, after which the rule ap-plier chooses what rule to apply . The then-part of the rule often creates new information, which is added to the working memory. This cycle is repeated until no more relevant rules are found .

Pros: An RBS is easy to understand, implement and maintain .

Cons: Since the solutions come from established rules, an RBS does not learn and does not im-prove its performance over time. An RBS can only be implemented if comprehensive knowledge is available .

Artificial neural networks (ANNs)

This type of AI uses, as the name suggests, a caricature of the way the human brain processes in-formation: an artificial neural network. It has many highly interconnected processing units working in unison . The network consists of an input layer, a number of hidden layers and an output layer. A neuron is linked to all neurons in the next layer. ANNs do not need a lot of prior assumptions, but learn from examples by adjusting the connection weights.

Pros: ANNs are useful in solving data-intensive problems where it is difficult to express the algo-rithm or rules to solve the problem . They are also robust to data errors.

Cons: They are, just like a CBR, a black-box model and therefore not suitable for problems that need process explanation .

Genetic algorithms (GAs)

This type of AI is a search technique mimicking natural selection . It evolves until the problem is satisfactorily solved: the fitter solutions in a population survive and pass their traits to offspring, replacing the poorer solutions. Successive populations are known as generations .

Pros: GAs are computationally simple and robust. They also balance load and efficacy well .

Cons: It can take a large computation time to get a result. Also, if a very fit individual emerges early and reproduces abundantly, early loss of diversity may lead to convergence on that local optimum .

Cellular automata (CA)

These are dynamic models which are discrete in time, space and state. The consist of a lattice of cells, interacting with their neighbours. The states of the cells are synchronously updated in time according to local rules. The new state of a cell at time t+1 is based on its state and those of its neighbours at time t .

Pros: Although CAs are simple mathematical models, they can simulate complex physical systems .

Cons: It is difficult to choose boundary conditions which reflect real life .

Fuzzy models

This type of AI uses fuzzy sets to deal with imprecise and incomplete data. Fuzzy set membership differs from normal set membership in that it takes a value between 0 and 1, instead of being true or false. This enables fuzzy models to describe vague statements as in natural language . Exact input values are transformed into fuzzy memberships through a process called fuzzi-fication . The model is then built on prior rules combined with fuzzified data by the fuzzy inference machine. The fuzzy output is then transformed to a crisp number, which is called defuzzification .

Pros: Fuzzy systems have a great ability to handle vague or imprecise information .

Cons: It is mostly more difficult to understand and apply than other AI techniques. Good member-ship functions are hard to determine. Also, fuzzy systems have no learning capability or memory .

Multi-agent systems (MAS)

This type of AI consists of a network of agents which interact to achieve goals . Each agent is a software component which contains code and data . Each agent on its own is incapable of solving the problem assigned to the MAS, but all agents together have the potential to solve it . The agents can communicate using a high-level Agent Communi-cation Language (ACL) through which they share information, request services and negotiate with each other .

Pros: MASs can model complex systems with multiple interactions among dynamic and autono-mous entities .

Cons: The effectiveness of a MAS depends largely on the agent organisation. When using peer-to-peer infrastructure, network maintenance can be problematic .

Swarm intelligence

This form of agent-based modelling is inspired by colonies of social animals such as bees and ants or schools of fish. This is interesting because simple individual agents can exhibit higher intelligence as a swarm. Local interactions can let global pat-terns emerge, without centralised control or a global model

Pros: Algorithms for SI techniques are versatile yet easy to implement. The group of agents has a self-organisation which allows adaptation to changes in the environment, which makes a SI system robust against failures and perturbations . This gives them the ability to solve dynamic problems as well .

Reinforcement learning (RL)

Reinforcement learning learns through interaction between a learning agent and the environment . Trial-and-error is used to achieve a goal.

Pros: RL is very useful in robotics and game playing , where it creates new behaviour rather than modelling existing behaviour.

Cons: RL alone is often not enough to solve other problems. RL in combination with other AI tech-niques is useful, but it is difficult to formulate a policy that works successful in real life .

Hybrid systems

In a hybrid system, two or more AI techniques are combined to overcome weaknesses presented by both techniques when used on their own. There are three main types of hybrid systems  :

• Sequential: the first technique passes its output to the second to generate the output.
• Auxiliary: the first technique obtains some information from the second to generate the output
• Embedded: the two techniques are contained within one another.

The most common hybrid is a neuro-fuzzy system, which combines an ANN and a fuzzy system. These are fast, efficient and easily designed, implemented and understood .

AI and simulations

Which AI technique to use for a simulation, depends on the case. If the process is complex and poorly understood, a black-box approach is favoured, for example by using CBR, ANN or GA. For problems with well-understood processes, RBS is great to apply. In this project, a couple of AI techniques can already be disregarded. Black-box approaches gives little insight in the system and processes involved. The aim of the simulation however is finding out which factors and which combination of factors has a positive influence on the revenue of a club or bar, which needs some insight in the system and processes involved. Therefore CBR, an ANN or a GA does not seem fit to use.

An RBS does seem fit to use. It is easy to understand, implement and maintain. The only thing needed is comprehensive knowledge of the situation and the patrons, which should be achieved using elaborate research.

AI articles not yet on the wiki

Research

Alcohol

Alcohol tolerance and the processing of alcohol by the body

One thing that will play an important role in the simulation, is alcohol and its effect on people. As is commonly known among many people, drinking alcohol makes one intoxicated, which can have all kinds of influences on people's behaviour . It is important to know which influences to correctly simulate the behaviour of a person over the night as they ingest more and more alcohol.

On top of that, each person has a limit to the quantity of alcohol they can ingest before they feel unwell and need to go home, which would be present in a realistic simulation as well. The amount of alcohol it takes before one feels drunk, differs per person. It depends among other factors on sex, size, and ethnicity of the person . A study into the number of drinks people needed before they considered themselves drunk found that African Americans reported to need less drinks before they felt drunk relative to whites. Hispanics reported to need more drinks than whites. These effects were stronger in the under 30 group . [NICE FIGURES IN PAPER] The same study found that on average, woman reported to need 4.15 drinks to feel drunk. Men needed 6.63 on average . It follows from this that when simulating drunkenness of people, each person needs to have a unique value for alcohol tolerance.

Intoxication can be observed, but it can also be measured through measuring the BAC, which estimates the degree of intoxication. In figure [x], the association between BAC ranges and the effects of alcohol in the average person are shown.

Fig [x]: General association between BAC ranges and the effects of alcohol in the average person

Of course ingested units of alcohol is not a direct indicator for someones BAC, contrary to what fig [x] suggests. For one, as already mentioned, every person reacts differently to alcohol. On top of that, over time the body does its best to process the alcohol. When ingesting alcohol, it of course goes through the stomach, where up to half of it is degraded before the remaining bits are passed into the small intestine. The amount of alcohol degraded in the stomach depends on the level of ADH in the person, which is generally higher in males than in females. Therefore men need more alcohol than women to feel drunk: more ADH means less drunk . A slower rate of absorption can be caused by a strong alcoholic drink on an empty stomach or the presence of fatty foods in the stomach. Also, carbonated alcohol or alcohol in combination with a carbonated drink will be absorbed faster than non-carbonated alcohol. Another factor which can affect the rate of absorption is the mood of the person . The absorbed alcohol is dilluted by body fluids. Therefore, larger people will have a lower BAC after ingesting the same amount than smaller people .

Thus, women, small people, and people who drink carbonated alcohol will become drunk faster.

A study found that pre-partying, drinking intentions and number of heavy drinking episodes are significantly associated with patron BrAC. Playing drinking games, patron race, student status, total drinking time, and plans to continue drinking did not seem to be indicators for BrAC . This suggests that a simulation may have to consider people already being a bit intoxicated when arriving at the bar, thus buying fewer drinks before they have reached their limit.

We also have to consider the time span in which the alcohol is ingested. About 15 ml of alcohol is metabolized per hour in an average-sized man . That is about one drink per hour. Thus there is a significant difference between drinking 6 units of alcohol in an hour and drinking 6 units of alcohol in 6 hours: the latter will give you a considerably lower BAC. Thus, if the simulation needs to be close to reality, a time factor should be factored in, where drunkenness goes down when not ingesting alcohol.

Because of its contents, alcohol can serve as nutrient and replenish energy. Per gram of alcohol, about 7.1 calories are released to the body .

Alcohol and behaviour

• Alcohol is reported to alter sensitivity to reinforcement and punishment

• Alcohol produces an effect that may be described as disinhibitory, related to an increase in behaviours that otherwise normally occur at a low rate. Many of these behaviours may be forms of risk-taking that result in aversive consequences to self or others: it was found that choices for the response option defined as risky were systematically increased as a function of alcohol dose .

Another study found that expected amount of alcohol consumed had an effect on risk-taking, but actual amount consumed did not.

• Multiple studies have been done on how alcohol intoxication influences impulsivity. This is often done by investigating degree of discounting, as degree of discounting correlates with tests on impulsivity . While one study found that alcohol had no effect on delay or probability discounting , another found that alcohol reduced impulsivity in that no-alcohol participants discounted delayed rewards at higher rates than intoxicated participants . An explanation for this contradiction could be that the first study completed the delay-discounting task five times, which possibly established a stable pattern of responding across the alcohol and placebo sessions, where the second study did not do that. In any case, it is clear that under certain conditions, alcohol intoxication reduces impulsivity.

It was also found that intoxicated participants were more likely to show lack of fit to the hyperbolic model, suggesting they respond less consistently .

Why people drink

Why people drink is important to find out, since this could give clues about what factors possibly have influence on the drinking behaviour of people.

Drinking gives a valuable opportunity to relax, have fun and form and maintain relationships. Young people ‘drink to get drunk’. Availability of cheap alcohol and drinks designed to attain drunkenness rapidly (such as shots and shooters) reinforce this norm of ‘drinking to get drunk’ in some contexts. Predrinking: the rightlevel of drunkenness was required to fully enjoy destination bars and clubs. Financial considerations are more important in limiting consumption than concerns about health. Many people restrict how much they drink by putting a monetary limit on the evening rather than one based on alcohol units or an idealised level of drunkenness or sobriety.

• Primary motivations for drinking include making it easier to socialize, loosen up, or open up .

• Becoming drunk could also be a way for young people to cope with the stresses and boredom of everyday life.

• Desires to drink and become intoxicated can also be framed around pleasure or fun.

Functions of pub-going

• Pub-going is associated with stronger social integration in peer networks. Visiting public drinking places is indicative of a way of life which facilitates social integration .
• Public drinking places offer the opportunity to meet the opposite sex and to start a romantic relationship

Conclusions

• Women feel drunk on average after 4.15 units of alcohol. Men feel drunk on average after 6.63 units of alcohol.
• About 15 ml of alcohol is metabolized every hour (around one unit)
• Per gram of alcohol, around 7.1 calories are released to the body.
• 10 ml of alcohol = 8 gram, so around 56.8 calories are released per hour (apart from the calories in the rest of the beverage, the not alcohol)
• Alcohol alters sensitivity to reinforcement and punishment
• Alcohol is disinhibitory
• Perceived alcohol intake increases risk-taking
• Under certain circumstances, alcohol intake reduces impulsivity
• People drink because they want to relax, have fun and maintain relationships.

Social structures

SOCIAL GROUPS EN WANNEER ZE GAAN DANSEN DUS MINDER ALCOHOL DRINKEN?

In-Pub behavioural data

• People who were in their local pub or community pubs were in significantly smaller social groups than those who were casual visitors in city centre bars. Those attending their local and those in community pubs were in conversation sized groups (maximum size 4): ; , whereas casual customers and those in city centre bars were typically in parties that were larger than the normative limit for conversations.
• The size of a social group had significant consequences for its dynamics. Conversations became more fragmented as the size of the group increased. And more people dropping out of the conversation as group size increased. The proportion of people who were not engaged with a conversation they were physically part of was significantly higher in city centre bars than in community pubs.

Social groups and their influence on the drinking behaviour of people in a bar

Drinking and becoming drunk is a highly sociable activity . Therefore one can imagine that the composition of ones drinking group can have quite an influence on how you drink. Several studies support that indeed, the social context of a drinking occasion can impact drinking behaviour . Several studies shed light on more specific social factors and their corresponding impacts.

For starters, a larger drinking-group size is associated with heavier drinking . This seems quite logical: with more people, there are more people doing rounds, thus there is a longer time between having to get alcohol yourself while still getting a continuous supply of alcohol. It has previously been found that when purchasing drinks in rounds, especially males tend to consume more alcohol .

The relationship between the number of friends and the number of drinks was stronger for men than for women , which corrolates with men drinking more when drinks are purchased in rounds. Also, men were found to consume more alcohol than women, particularly at the beginnings of the evening. .

On top of the number of friends you take with you, the gender composition of those friends also influences drinking behaviour. A study found that both males and females consume significantly more drinks in mixed-gender groups. . There is a difference between males and females here: men consumed more drinks in groups with an equal amount of males and females and groups with men in the majority. They also consumed more drinks with in a group with men only compared to woman only.

Woman however, although they did consume more drinks in mixed-gender groups, consumed significantly fewer drinks in groups with men only than groups with woman only.

There are more differences between male and female pub-goers. One study found that male friends or acquaintances were the main sources of pressure on people to drink or drink more . According to the same paper, pressure to drink also depends on religion and gender. Another study found that indirect pressures to drink play a more significant role than direct pressures .

Peer pressure

Drinkers with companions who consumed large amounts of alcohol tended to consume more alcohol and tended to have higher drinking rates .

Conclusions

• Larger drinking group size => more alcohol. Stronger relationship for men than for women.
• Men consume more alcohol than women, particularly at the beginning of the evening
• Both men and women like mixed-gender groups the most to drink in. People hate drinking with only men.
• Men are the main source of pressure to drink more
• Peer pressure is a thing in alcohol drinking

Music and its influence on the behaviour of people in a bar

Music can have quite an influence on how people behave in social settings. Most people like background music: studies have found that people spend less time drinking in bars that don’t play music . It goes even further: according to a previous study, loud music makes people drink 31% more . The reason behind this is could be that when people can not communicate due to the noise in the bar, they start focussing on drinking . Another explanation is that high sound levels may cause higher arousal, which leads the subjects to drink faster and order more drinks .

Another study links faster music to faster drinking . However, a different study links slower music to more sales. Customers stayed about the same amount of time but spent more during slow music than during normal or fast music. . An explanation for this difference is that people prefer to listen to music that moderates their state of arousal . During a relaxing activity like an afternoon beer, people prefer slow low-arousal music, while during a night out people prefer music that further heightens their state of arousal. On top of that, studies found that the genre of music matters. Jacob (2006) found that when playing drinking songs in a bar, the duration of stay and spending both are increased . Further research supported this and claimed that customers exposed to textual references to alcohol spent significantly more on alcoholic drinks than those who were not .

It was also found that in social context, people modify their bodily behaviour according to the dynamic level of the bass drum. More specifically, they move more actively and display a higher degree of tempo entrainment as the sound pressure level of the bass drum increase . This could be interesting if a correlation between dancing and alcohol could be found.

Conclusions

• Background music is a plus
• Louder music => more drinks
• Faster music => faster drinking, but not during a relaxing after noon beer, then slower music => more beer
• Drinking songs => more drinks
• more bass => more dancing

Influence of bar-related factors

• Day of the week: heavy drinking occurs mostly on Saturday evenings followed by Friday evenings. This is because young people do not have any responsibilities the next day

Scent

Apart from the earlier named direct influences on alcohol intake, a lot of other factors can influence how people behave on a night out.

One of these factors is scent: on a night out, one may encounter many scents, both pleasant and vile. Sweating bodies, alcohol, you name it. It would seem logical that ambient scents that mask the vile odours could contribute positively to the night-time experience. A previous study confirms this: they tested the scents of orange, seawater and peppermint, and found that all scents enhanced dancing activity and improved the evaluation of the evening, music and mood . Additionally, they found that the increase in dancing coincided with an increase in temperature in the club. This may lead to the customers wanting more alcohol: if it is hot, what’s better than a cold beer to cool off?

Factors that influence how much people like a bar

Apart from knowing how to influence the alcohol intake of customers, it is important to know how to get customers. There have been a few studies that worked with focus groups in order to find out what people wanted from bars.

First of all we have security. Most bars have security, if only to dismiss all people below legal drinking age. The nature of the security presence outside a bar can be a good indicator of the level of security inside a bar. Studies found that a more formal attire offers a greater image of security . This is because people do not like going somewhere where you do not feel safe. People however wanted the security personal to look friendly as well, they did not want to feel threatened by the very security that should protect them . There is a difference here between male and female clientele: while most men belief that more formal attire offers the greatest image of security, only about 50% of woman have the same belief, while the other 50% prefer an informal attire.

The clientele is important as well. People prefer a mix of male and female clientele, with the majority liking a male-dominated place the least . This seems to corollate with the fact that people drink more in mixed-gender groups. Women like a mixed clientele more than men; men like a predominantly female clientele about as much as a mixed clientele . Males offered an explanation for their preference for mixed-gender bars: they identified woman as critical factor in the decision to select a bar . How busy it looked is also named as a critical factor in the decision to select a bar. It should not be too busy, but not to quiet either .

Then we have the seating. Opposing the stereotypical image of pub customers on barstools, the same study found that sofas are the most preferred as seating arrangement inside a venue. Seating and bar stools proved far less popular. People like relaxing and taking a break from the pushing and shoving every now and then . Woman like individual seating particularly well, but still less than sofas .

The type of venue was found to be important as well. Both men and women prefer a traditional bar, but men prefer this more than women. Women have as close second the wine bar. A Latin themed bar is least liked by both men and women .

Lastly, the alcohol prices are important for choosing a bar, but not as much as other factors. People like everyday low alcohol prices but are prepared to pay a bit more, as long as it is not too much . Low alcohol prices do have a significant effect on alcohol intake however: another study found that a happy hour with price reduction increases alcohol consumption. When the purchase price was reduced by half, casual and heavy drinkers increased their consumption eight and nine times respecively .

Preferred location of the dance floor : 1. Surrounded on all sides by people 2. Away from the bar 3. near the bar 4. Surrounded on two sides by people

Conclusions

• A nice scent in the pub ==> better evening
• Security: For male formal, for female does not matter
• Mix of male and female clientele
• Not too quiet, not too busy
• Sofas > individual seating > bar stools
• Low alcohol prices => more alcohol intake

Activities and alcohol consumption

Available activities in bars, like playing games, watching TV or making conversation, has influence on the alcohol intake of people as well. A study found that especially in males, active pastime activities like playing pinball, playing cards or playing table football, result in slower drinking than passive pastime activities like being alone, making conversation or watching TV . The males in the study displayed the same drinking rate as women when active, but a slower one when passive. However they compensate for this 'lost time' during the passive activity following the active one by drinking more. They also do drink more alcohol during conversation than females .

Why people do certain activities

People dance because:

• Strongest motivational factor is mood enhancement, followed by self-confidence
• Women dance for reasons of fitness, mood enhancement, trance, self-confidence and escapism than men.
• Men were mostly motivated by intimacy.
• No significant difference regarding socialising and mastery.

Many of the motivational factors related to drinking alcohol also appear in the motivations for dancing, such as mood enhancement, socialising, and escapism. However self-confidence and intimacy are specific to dancing .

Most people dance for enjoyment, and prefer dancing with good friends over dancing with someone they are sexually interested in.

Conclusions

• Passive activities = more alcohol intake than active activities
• After active activities, males compensate for their "lost time" by drinking slightly faster.

Other

Nog te verwerken

Approach

1.Determine important variable for simulation

2. Research state of the art.

2.1 Determine correlations between the determined variables and alcohol consumption trough state of the art.

2.2 formalyze hypothesis etc.

3.Determine unknowns in correlations, research those correlations.

3.1 Set up a research plan.

3.2 Execute the research plan.

3.3 Analyze the results.

4. Set up a simulation using the found correlations.

4.1 Create a minimum viable product: a bar setting with AI using that bar.

4.2 start implementen each of the correlations found in 2 and 3.

5 Analyze the simulation

5.1 Change variables, optimize the simulation.

5.2 Refer back to research and hypothesis, is our simulation realistic and how does it comply with our hypothesis?

5.3 If neccecary review steps 2, 3 and 4.

6 Finalize wiki and conclusions

Planning

• Week 1: research state-of-the-art, finalise plan
• Week 2: More research, state requirements for simulation, create research plan for filling in the blanks of the state-of-the-art.
• Week 3: Implemented first version of simulation with only basic features
• Week 4: Implemented second version simulation with requirements implemented
• Week 5: Performing simulation, documenting results
• Week 6: Performing simulation, documenting results
• Week 7: Compare results to real life, create conclusion
• Week 8: Finalise wiki

Gantt chart

Milestones

Tasks	Estimated Time
Planning	12:00
State-Of-The-Art research	75:00
Conclusions state of the art	5:00
Set up research plan	5:00
Execute research plan	50:00
Analyze results	20:00
Minimum viable product	50:00
Simulation with more complex AI	100:00
Analyze simulation	75:00
Update research simulation based on results	50:00
Write down conlcusions	20:00
Finalize wiki	20:00
Create presentation	10:00
Unforseen	100:00
Total	592:00

Deliverables

State-of-the-art

A list of variables influencing behaviour at a bar setting.

A list of missing variables influencing behaviour at a bar setting.

A hypothesis on how the results our simulation would create.

A research plan detailing how to fill in the blanks of the state of the art.

Results of our research.

A minimum viable product: a simulation that can houses AI that can navigate a bar setting.

Extension of the minimum viable product trough implementing more complex behaviours found in the research.

Results out of the simulation, a set of factors and the behaviour it creates in the simulation.

Finalized wiki and conclusions.

A presentation

Who is doing what

Week 1

Name	Total	Break-down
Daan	2h	Discussing the subject
Job	2h	Discussing the subject
Sanne	5h	Making a draft for the wiki (0.5 h), Gathering links for the State of the Art (2.5h), Discussing the subject (2h)
Jasper	3h	Gathering articles for state-of-the-art (1.5 h), Discussing the subject (2h)
Wietske	6h	Working on the wiki (0.5 h), Gathering articles for state-of-the-art (3.5 h), Discussing the subject (2h)

Week 2

Name	Total	Break-down
Daan	10h	Working on the wiki (3 h), gathering articles for state-of-the-art (3 h), Discussing the subject (4h)
Job	10h	Working on the wiki (2 h), gathering articles for state-of-the-art (4 h), Discussing the subject (4h)
Sanne	10h	Working on the approach and planning (3h), working on the state of the art (3h) Discussing the subject (4h)
Jasper	10h	Working on the wiki (2 h), gathering articles for state-of-the-art (4h) Discussing the subject (4h)
Wietske	11.5 h	Working on the wiki (1 h), discussing the subject (4h), gathering articles for state-of-the-art (3 h), organizing sources on wiki by subject (1h), writing about the effect of music (1h), writing about social groups and alcohol(1.5 h)

Week 3

Name	Total	Break-down
Daan	14h	Working on the wiki (7h), gathering articles for state-of-the-art (3), Discussing the subject (4h)
Job	12h	Working on the wiki (6h), gathering articles for state-of-the-art (2h), Discussing the subject (4h)
Sanne	4h	Working on the approach and planning (), working on the state of the art () Discussing the subject (4h)
Jasper	11h	Working on the wiki (1h), Discussing the subject (4h), setup Github(0.5h), Started programming simulation (5.5h)
Wietske	18h	Working on the wiki (2 h), discussing the subject (4h), gathering articles for state-of-the-art (1 h), writing about factors in a bar (2 h), incorporating articles in state-of-the-art (2h), reading articles (4h), writing about alcohol (3h)

Week 4

Name	Total	Break-down
Daan	26h	Discussing the subject (4h), improve/create code skeleton (4h), Alter data structure for more complete result collection (4h), create simulation flow + implement more club variables(2h), Experimented with exporting variables to MATLAB (2h), Create visualisation for crowd and bar objects (3h), Create crowd position/movement system (3h), Create collision system for movement (3h), implement alcohol vending and consumption (1h)
Job	14h	Discussing the subject (4h), gathering articles for state-of-the-art (4h), working on the wiki (2h), working on java for Matlab (4h)
Sanne		Discussing the subject (4h)
Jasper		Discussing the subject (4h), Making small changes in the code (1h),
Wietske	16h	Discussing the subject (4h), Working on the wiki (2h), writing about alcohol (2h), Reading and processing articles (4h), writing about AI (2h), organising state-of-the-art (1h), contacting bar owners (1h)