replace words like therefore with synonyms

30_Thesis/sections/60_evaluation.tex

@@ -293,7 +293,7 @@ When looking at dissatisfaction change the graphs are all in the negative number
 The figures for homogenous (\autoref{fig:Evaluation:HomoSatisfaction}) and random groups (\autoref{fig:Evaluation:RandomSatisfaction}) have a similar shape but their values and slopes vary. The satisfaction change for homogenous groups is mostly negative, starting at $-2$, and only reaches a positive level for more than $100$ stored configurations with a value of $0.04$. Multiplication and best average have higher values than least misery here too. Moreover the dissatisfaction change is always positive with a value range of $[0,1]$, except for a slight fall below zero after more than $75$ configurations are stored.
 Random groups as seen in \autoref{fig:Evaluation:RandomSatisfaction} mostly have a positive change in satisfaction. Here, values range from $-0.55$ to $0.27$ for least misery, from $-0.27$ and $-0.28$ to $0.74$ for best average and multiplication. The change is higher than the change for heterogeneous groups. Dissatisfaction also changes similarly to heterogeneous groups. Here the values for random groups reach a lower level. They range from $0$ to $-0.59$ for least misery. Multiplication and best average both have a minimum value around $-0.21$ and behave similarly. The range goes down to $-0.84$ for best average and $-0.86$ for multiplication.
 
-\autoref{fig:Evaluation:HeteroSatisfaction} also shows the average number of group members satisfied and dissatisfied with the recommender's decision. Satisfaction with the recommender's decision starts at $2.4$ and quickly reaches $2.65$ for least misery and $2.8$ for best average and multiplication. The highest value for multiplication is at $2.89$. Dissatisfaction also quickly plateaus. Here values for different recommenders are closer together. They start at $0.74$ (least misery) to $0.78$ (best average) and fall as low as $0.62$ for least misery, $0.66$ for best average and $0.56$ for multiplication.
+\autoref{fig:Evaluation:HeteroSatisfaction} also shows the average number of group members satisfied and dissatisfied with the recommender's decision. Satisfaction with the recommender's decision starts at $2.4$ and quickly reaches $2.65$ for least misery and $2.8$ for best average and multiplication. The highest value for multiplication is at $2.89$. Dissatisfaction likewise quickly plateaus. Here values for different recommenders are closer together. They start at $0.74$ (least misery) to $0.78$ (best average) and fall as low as $0.62$ for least misery, $0.66$ for best average and $0.56$ for multiplication.
 
 When looking at the total numbers as shown in \autoref{fig:Evaluation:HomoSatisfaction} the value range for homogenous groups is much larger but the overall shape stays the same. Here satisfaction numbers go from $0.55$ to $2.95$. Least misery performs visibly worse than multiplication and best average reaches only $2.7$. Dissatisfaction values range from $1.21$ to $0.01$ and the values are not really visibly distinguishable, except in the range of $[25,50]$. Least misery seems to have the highest number of dissatisfied group members in this range.
 
@@ -304,10 +304,10 @@ Random groups have less overall satisfaction with $tc = 85\%$ as seen in \autore
 Having described the data the remaining hypotheses are discussed.
 \autoref{hyp:Evaluation:HomogenousMoreSatisfied} states that homogenous groups have more satisfied members with regards to the dictator's and the group recommender's decision. \autoref{fig:Evaluation:tcCount} shows that this holds true for the dictator's decision as for every instance satisfaction in homogeneous groups is higher than that of other groups. However, \autoref{fig:Evaluation:HeteroSatisfaction}, \autoref{fig:Evaluation:HomoSatisfaction} and \autoref{fig:Evaluation:RandomSatisfaction} show that for satisfaction with the recommender's decision this does not hold true when looking at $tc$ values where the recommender performs best for each segment. In these cases the homogenous group only reaches the highest amount of satisfaction when the recommender has access to all stored configurations. With a decreasing number of stored configurations both random groups and heterogeneous groups achieve a higher satisfaction. This is likely to happen due to the similarity between group members. A recommender with imperfect knowledge and a size-reduced, configuration database generates results that are not good enough and cannot compete with the dictator who always finds the perfect individual match that group members of homogeneous groups are satisfied with. It is important to note that homogeneous groups show a higher number of satisfied people across the board when the same $tc$ values are used.
 
-\autoref{hyp:Evaluation:HeterogenousBiggerSatisfactionIncrease} states that the increase in satisfaction should be bigger for more heterogeneous groups. However, \autoref{fig:Evaluation:HeteroSatisfaction}, \autoref{fig:Evaluation:HomoSatisfaction} and \autoref{fig:Evaluation:RandomSatisfaction} show this not to be true. The recommendations for heterogeneous groups indeed cause a larger change in satisfaction compared to homogeneous groups but random groups cause a positive change of higher magnitude. Also, the decrease in dissatisfaction is higher among random groups. This possibly happens because random groups have more aligned interests and preferences among group members and, therefore, they do not diverge as much which results in compromises for the group that can satisfy more individual members. Also the group preferences are still far enough apart to cause dissatisfaction and neutrality with the dictator's decision.
+\autoref{hyp:Evaluation:HeterogenousBiggerSatisfactionIncrease} states that the increase in satisfaction should be bigger for more heterogeneous groups. However, \autoref{fig:Evaluation:HeteroSatisfaction}, \autoref{fig:Evaluation:HomoSatisfaction} and \autoref{fig:Evaluation:RandomSatisfaction} show this not to be true. The recommendations for heterogeneous groups indeed cause a larger change in satisfaction compared to homogeneous groups but random groups cause a positive change of higher magnitude. Also, the decrease in dissatisfaction is higher among random groups. This possibly happens because random groups have more aligned interests and preferences among group members and, therefore, they do not diverge as much which results in compromises for the group that can satisfy more individual members. Likewise, the group preferences are still far enough apart to cause dissatisfaction and neutrality with the dictator's decision.
 
-The data shows that a larger configuration database causes the amount of satisfied group members to be greater than recommendations using a smaller database. With dissatisfaction the same is seen in the inverse. A larger configuration database causes the number of dissatisfied group members to drop compared to a small database. However, in some runs there have been instances of least misery that show a slight drop. This can be seen in \autoref{fig:Evaluation:HeteroSatisfaction} when comparing $74$ and $148$ as the number of stored configurations. Why this happens is not entirely clear but a cause might be that least misery just takes into account the worst performing group member of the group. Therefore, it is possible that there is a second slightly worse solution, when comparing least misery scores, which actually has a slight advantage in terms of dissatisfaction. This second best configuration can cause it to be in the second database partition therefore resulting in less dissatisfaction on average. \autoref{hyp:Evaluation:StoreSizeBetterResults} therefore is supported by the data but it does not fully hold true when looking at least misery.
+The data shows that a larger configuration database causes the amount of satisfied group members to be greater than recommendations using a smaller database. With dissatisfaction the same is seen in the inverse. A larger configuration database causes the number of dissatisfied group members to drop compared to a small database. However, in some runs there have been instances of least misery that show a slight drop. This can be seen in \autoref{fig:Evaluation:HeteroSatisfaction} when comparing $74$ and $148$ as the number of stored configurations. Why this happens is not entirely clear but a cause might be that least misery just takes into account the worst performing group member of the group. Thus, it is possible that there is a second slightly worse solution, when comparing least misery scores, which actually has a slight advantage in terms of dissatisfaction. This second best configuration can cause it to be in the second database partition hence resulting in less dissatisfaction on average. \autoref{hyp:Evaluation:StoreSizeBetterResults} thus is supported by the data but it does not fully hold true when looking at least misery.
 
-\autoref{hyp:Evaluation:AggregationStrategies} states least misery performs worse than multiplication. For a change in satisfaction this can be seen across the board, however, for the change in dissatisfaction this is not true everywhere. \autoref{fig:Evaluation:HeteroSatisfaction} shows that least misery performs better than best average in terms of dissatisfaction reduction. This behaviour possibly occurs because an average metric yields the same results for heavily polarised decisions and decisions that everyone feels neutral about. Least misery on the other hand takes into account only the group member least satisfied with the decision and, therefore, this metric performs better. However, in other cases it performs visibly worse. Also notable is that multiplication performs best across the board. This supports the findings by \citeauthor{Masthoff2015} \cite[~p. 755f]{Masthoff2015} and also shows that the satisfaction model does show some similar results to online evaluations.
+\autoref{hyp:Evaluation:AggregationStrategies} states least misery performs worse than multiplication. For a change in satisfaction this can be seen across the board, however, for the change in dissatisfaction this is not true everywhere. \autoref{fig:Evaluation:HeteroSatisfaction} shows that least misery performs better than best average in terms of dissatisfaction reduction. This behaviour possibly occurs because an average metric yields the same results for heavily polarised decisions and decisions that everyone feels neutral about. Least misery on the other hand takes into account only the group member least satisfied with the decision and, therefore, this metric performs better. However, in other cases it performs visibly worse. Also notable is that multiplication performs best across the board. This supports the findings by \citeauthor{Masthoff2015} \cite[~p. 755f]{Masthoff2015} and further shows that the satisfaction model does show some similar results to online evaluations.
 
 To go back to the evaluation questions posed in \autoref{sec:Evaluation:Questions} this section has shown that for random and heterogeneous groups the recommender performs better than a dictator. The average satisfaction depends on the chosen parameters but for the chosen value range average satisfaction with the recommender decision lies above two and can reach close to three satisfied group members for a high number of stored configurations and for some group types. The amount of stored finished configurations plays an important role in the recommender's performance but with a fraction of stored configurations the recommender still yields good results. This shows that the recommender provides useful decision support for helping in group decisions. It provides a solid basis for groups and can help their group decision. Most decisions the recommender does improve group satisfaction which shows that the recommender is able to be used to improve group decisions.

30_Thesis/sections/80_conclusion.tex

@@ -20,12 +20,12 @@ Due to the scope of this thesis it was not possible to analyse all possible scen
 
 An offline satisfaction metric is introduced in this thesis and it could lead to results that differ from real life impressions of people. It has yet to be validated in a real-world setting. The validations of this metric allows the use of another metric for scenarios that lack a suitable existing metrics. Moreover, understanding ´the relation between the introduced metric and actual satisfaction can help to form more accurate satisfaction models. This helps to understand finding better compromises and such a metric can directly be used as a component of a group recommender.
 
-In this thesis only one use case was considered. Therefore, different use cases could yield different results, in terms of satisfaction and the tolerance of limited knowledge. It is unclear if different scenarios that either are more complex or greater in size yield the same results. The successful application to larger sized products could help for large complex projects that have multiple experts from differing areas that do not necessarily agree. Also, larger products potentially reach limitations of the recommender as the solution space grows quickly. An approach that clusters configurations and other means of optimization can help with performance. Complex and large products might lead to usability issues as users potentially are overwhelmed with choices and setting all preferences individually might be infeasible. Therefore, indirect means of acquiring preferences should be incorporates for more complex and larger configurations. 
+In this thesis only one use case was considered. Therefore, different use cases could yield different results, in terms of satisfaction and the tolerance of limited knowledge. It is unclear if different scenarios that either are more complex or greater in size yield the same results. The successful application to larger sized products could help for large complex projects that have multiple experts from differing areas that do not necessarily agree. Also, larger products potentially reach limitations of the recommender as the solution space grows quickly. An approach that clusters configurations and other means of optimization can help with performance. Complex and large products might lead to usability issues as users potentially are overwhelmed with choices and setting all preferences individually might be infeasible. Thus, indirect means of acquiring preferences should be incorporates for more complex and larger configurations. 
 
-Furthermore, groups were automatically generated and thus, possibly lack resemblance to common real world group constellations. Therefore, real world group constellations for decisions should be analysed. This allows the validation of the implemented groups and can lead to a better understanding of generating synthetic groups that resemble actual groups. Thus, less time has to be spend in acquiring user data.
+Furthermore, groups were automatically generated and thus, possibly lack resemblance to common real world group constellations. Consequently, real world group constellations for decisions should be analysed. This allows the validation of the implemented groups and can lead to a better understanding of generating synthetic groups that resemble actual groups. Thus, less time has to be spend in acquiring user data.
 
-Moreover, the recommender performed not ideal with homogenous groups especially when knowledge about the solution space was limited. Therefore, methods of detecting homogenous groups could detect cases in which the recommenders perform badly and use other recommenders instead.
+Moreover, the recommender performed not ideal with homogenous groups especially when knowledge about the solution space was limited. Hence, methods of detecting homogenous groups could detect cases in which the recommenders perform badly and use other recommenders instead.
 
 Additionally, the group size was fixed to four members and deviating results for differently sized groups are possible. Moreover, this approach can be extended to potentially allow a whole community of hundreds of people to decide about neighbourhood changes. This could range from the layout of a new community centre, staffing, equipment and uses. Therefore, such approaches of group-based configuration can be used for public participation in projects helping communities to build trust and be more involved in decisions.
 
-Finally, the approach used in this thesis assumes a flat group hierarchy. Modelling knowledge and hierarchy of a group can help to improve group decisions further as supervisors do not feel overrun by their employees and knowledge of experts on certain parts of a product or solution can use that knowledge to guide the decision that area. Experts in other areas could have more say in areas of their expertise. Therefore, decisions could be expert and hierarchy driven which should help with group satisfaction about compromises.
+Finally, the approach used in this thesis assumes a flat group hierarchy. Modelling knowledge and hierarchy of a group can help to improve group decisions further as supervisors do not feel overrun by their employees and knowledge of experts on certain parts of a product or solution can use that knowledge to guide the decision that area. Experts in other areas could have more say in areas of their expertise. Thus, decisions could be expert and hierarchy driven which should help with group satisfaction about compromises.