Update recommender readme.md (#572)

# Description


Closes #XXX

## How to Test


## Checklist
- [ ] The code includes tests if relevant
- [ ] I have *actually* self-reviewed my changes and done QA

apps/algorithm/recommend/readme.md

@@ -24,17 +24,18 @@ The recommender is built using TensorFlow, with its API in FastAPI.
 
    If you would like to see the integration with frontend, you can navigate to
    the homepage at `http://localhost/`. Once you make an account, you can see
-   the default recommendations. Perform a search using the search bar. Note that
-   our dataset is limited, so you may want to stick with items we have (some
-   good terms include chocolate, hat, perfume, etc). Then, navigate back to the
+   the default recommendations. Perform a search using the search bar. Then, navigate back to the
    homepage, and notice that some of your recommendations have changed. The term
-   you searched may appear in the title of these items, or their descriptions.
-   It's also possible that searches may be flaky, as Elasticsearch likes to fail
-   for reasons currently unknown. If you're getting `502` errors while hitting
-   the search endpoint, let us know and we can just demo it for you in lab.
-
-   Note that the recommender works for item clicks as well, but we have no way
-   for you to test this. We can demonstrate this to you during a lab if necessary.
+   you searched may appear in the title of these items, or their descriptions. Next, you can test
+   recommendations based on clicks. If you're looking at the recommendations page and you click on
+   an item, then you can go back to the homepage and see how the recommendations change by showing
+   more of the same item you just clicked on. Finally, you can look at your recommended items, and
+   choose that you're not interested in the given item. If you refresh your homepage, this recommendation
+   will disappear along with "similar" recommendations, assuming the recommendation strength of the
+   your searches and clicks doesn't overpower the similar not interested items, since not interested
+   items are negated whereas searches and clicks are additive. For example, if you search up "banana"
+   many times and click on many banana products, and choose to stop suggesting items similar to a
+   banana product, chances are you will still see some banana products. 
 
 ## Testing Instructions