Basic question: Determine how significant a deviation is I'm just getting started with statistics - so far I'm familiar with the basics of descriptive statistics. It's a basic question that has probably been answered a thousand times. However I don't know what keywords to put in to read about it.
I want to make a quantitative statement that proves the hypothesis that there's a gap/ negative relationship between blue (attitude) and red (behavior). Also I want to make separate statements for each category (strongly agree, agree, neutral, ...)
Here's a picture of the survey:
n = 52
both categories (red, blue) were separate questions, meaning no AND-questions

 A: You didn't post the raw data, but I will assume that you have that available.  In other words, I will assume that you have raw data of the form
+-----------+---------------------+--------------------+
| person_id | concerned_about_env | look_for_ecolabels |
+-----------+---------------------+--------------------+
|         1 | strongly agree      | strongly agree     |
|         2 | agree               | agree              |
|         3 | neutral             | agree              |
+-----------+---------------------+--------------------+

If you have this level of detail, then you could use Kendall's Tau as your test statistic.  Kendall's Tau can be used to measure the association between two ordinal variables.  For your case, you could use Kendall's Tau to measure the association between attitude and behavior.
Your null hypothesis would be that there is no association between attitude and behavior ( = 0) while your alternative hypothesis would be that there is a negative association ( < 0).
Example R code:
library(ggplot2) # optional for plotting data

set.seed(123) # for reproducibility
# Use coding:
# 5 = strongly agree
# 4 = agree
# 3 = neutral
# 2 = disagree
# 1 = strongly disagree

# Make up data
concerned_about_env = rep(seq.int(5, 1, -1), times = c(28, 20, 3, 1, 0))
look_for_ecolabels = rep(seq.int(5, 1, -1), times = c(4, 22, 11, 11, 4))

concerned_about_env = sample(concerned_about_env, 
                             size = length(concerned_about_env),
                             prob = seq_along(concerned_about_env)/length(concerned_about_env))

look_for_ecolabels = sample(look_for_ecolabels, 
                             size = length(look_for_ecolabels),
                             prob = rev(seq_along(look_for_ecolabels))/length(look_for_ecolabels))

# Plot data (optional)
ggplot(data = data.frame(concerned_about_env, look_for_ecolabels),
       aes(x = concerned_about_env, y = look_for_ecolabels)) +
  geom_jitter(width = 0.1, height = 0.1, alpha = 0.5)

# Test the null hypothesis that Kendall's tau is 0 vs
# the alternative hypothesis that Kendall's tau is less than 0.
test1 = cor.test(x = concerned_about_env,
         y = look_for_ecolabels,
         alternative = "less",
         method = "kendall",
         continuity = TRUE,
         exact = FALSE)

print(test1)

You can follow this up by looking at which cells in the contingency
table differ from their expected values.
contingency_table = table(concerned_about_env, look_for_ecolabels)

test2 = chisq.test(contingency_table)

o = test2$observed
e = test2$expected
n = sum(o)

adjusted_residuals = (o - e)/sqrt(e * (1 - rowSums(o)/n) * (1 - colSums(o)/n))

For Fisher’s exact approach for post hoc analysis of a chi-squared test, see code and paper.
