\section{Conclusion}
\label{sec:conclusion}
Since keyboards are still the main data input device while using a computer and
are also found to be related to discomfort or even \glsfirst{WRUED}
\cite{pascarelli_wrued}, we tried to evaluate a possible modification to the
existing keyboard design, that does not require the consumers to extensively
adapt their typing behaviour, nor the producers to massively deviate from
existing manufacturing processes. To reduce the load on weaker fingers, we
created a keyboard that used keyswitches with actuation forces, related to the
specific finger the keyswitch is operated with and hoped to thereby decrease the
risk for \gls{WRUED}. The evaluation of the impact of different actuation forces
on typing speed, error rate and satisfaction revealed, that higher actuation
forces reduce error rates compared to lower actuation forces and that the typing
speed is also influenced―\textbf{at least indirectly}―by differences in
actuation force. Especially the keyboard with very low actuation force,
\textit{Nyx (35\,g)}, which also had the highest average error rate was
significantly slower than all other keyboards. Therefore, we investigated, if
there is a connection between high error rates and stagnating typing speed and
found that in general, the error rate was a factor for lower input
rates. Neither the satisfaction nor the muscle activity was significantly
influenced solely by the actuation.

In conclusion, our study showed that the keyboard with non-uniform actuation
forces―\textit{Hera (35 - 60\,g)}―was not able to improve the overall typing
experience significantly enough to supersede existing designs, but was still a
viable alternative to all traditional keyboards we tested. It could be possible
that due to the unconventional force distribution, that is similar to keyboards
with very light actuation force, the muscle activity while using \textit{Hera}
could decrease when users are given more time to adapt to this keyboard
\cite{gerard_keyswitch}.  Additionally, we found that keyboards with either very
high (80\,g) or very low (35\,g) actuation forces had the most influence on
typing related metrics, when compared to the more commonly sold keyboards with
around 50\,g to 60\,g actuation force. In the next sections we identify possible
limitations and propose some ideas on how to reevaluate custom keyboard designs
in future studies.

\subsection{Limitations}
\label{sec:limitations}
The first limitation of our study design was the rather short time period of in
total 10 minutes, for every participant to adjust to each keyboard. With
prolonged typing session, familiarization, especially for keyboards with lighter
actuation force, would have been more realistic to a real life scenario where a
person bought a new keyboard. Furthermore, the laboratory test environment where
the researcher was in the same room, the limited time for the individual typing
tests and the rather short breaks in between typing tests, could have influenced
some subjects by inducing unnecessary stress. Another limitation related to the
preliminary finger strength study, was the very small number of participants (n
= 6). Although we measured the finger strengths in different positions for
50\,\% female and male participants, the age distribution was not diverse (M =
24) and with a higher number of subjects, the results would have been much more
reliable. Similarly, the number and diversity in occupation of participants
could have been higher for our main study (n = 24) to yield even more meaningful
results. The low number of participants in general was partly due to the ongoing
COVID-19 pandemic. Lastly, we could have used more linear mixed models during
our statistical analysis, to be able to make statements about the influence of
other factors e.g., age, gender, average daily keyboard usage, etc., on speed,
error rate or satisfaction.

\subsection{Future work}
\label{sec:fw}
We propose, that in further research related to keyboards with non-uniform
actuation force (adjusted keyboards), participants should test several different
adjusted keyboards and the results should be compared to one identical looking
keyboard that utilizes a uniform layout of keyswitches with an actuation force
of 50\,g to 65\,g. Further, different adjusted layouts, with e.g. higher or
lower base actuation force than 50\,g could be used to calculate the individual
spring resistances used for each key or a similar layout to the one used in
Realforce\footnote{\url{https://www.realforce.co.jp/en/products/}} keyboards,
could be compared to each other. Furthermore, long term studies with adjusted
keyboards where participants use the adjusted keyboard for three to four months
and then use a uniform keyboard they prefer for another three to four months as
their daily driver, could yield more accurate results, due to the chance to
fully adapt to the individual keyboards. During those months \gls{EMG} and
typing related metrics should be measured on a regular basis. Lastly, it would
be interesting to investigate if an adjusted keyboard can reduce pain or at
least enhance comfort for typists with pre-existing diseases influenced by
typing activities (disorders of the upper extremity), since one of our
participants with a similar disease reported a great reduction in pain while
using \textit{Hera}.